WO2023051312A1 - 一种道路决策方法、***、设备和介质 - Google Patents

一种道路决策方法、***、设备和介质 Download PDF

Info

Publication number
WO2023051312A1
WO2023051312A1 PCT/CN2022/119830 CN2022119830W WO2023051312A1 WO 2023051312 A1 WO2023051312 A1 WO 2023051312A1 CN 2022119830 W CN2022119830 W CN 2022119830W WO 2023051312 A1 WO2023051312 A1 WO 2023051312A1
Authority
WO
WIPO (PCT)
Prior art keywords
waypoint
target
unmanned vehicle
value
lane
Prior art date
Application number
PCT/CN2022/119830
Other languages
English (en)
French (fr)
Inventor
李思思
张賾隐
韩旭
Original Assignee
广州文远知行科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN202111155395.9A external-priority patent/CN113911136B/zh
Priority claimed from CN202111453193.2A external-priority patent/CN113928341B/zh
Application filed by 广州文远知行科技有限公司 filed Critical 广州文远知行科技有限公司
Publication of WO2023051312A1 publication Critical patent/WO2023051312A1/zh

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W60/00—Drive control systems specially adapted for autonomous road vehicles

Definitions

  • the present application relates to the technical field of unmanned vehicles, and in particular to a road decision-making method, system, device and medium.
  • Unmanned vehicle is a comprehensive intelligent platform integrating multiple functions such as environmental perception and cognition, dynamic planning and decision-making, behavior control and execution.
  • the core issues of unmanned vehicle research include environmental perception, behavior decision-making and motion control.
  • Road decision-making is the main component of unmanned vehicle decision-making technology.
  • the existing technology usually adopts the method of model predictive control to obtain the optimal road decision, and obtains the optimal road decision by solving complex optimization problems, which requires a lot of computing power to solve Non-linear optimization problems lead to low efficiency of road decision-making, and it is difficult to be effectively applied to the decision-making system of unmanned vehicles.
  • This application provides a road decision-making method, system, equipment, and medium, which are used to improve the prior art by adopting model predictive control methods to obtain optimal road decisions.
  • Obtaining optimal road decisions by solving complex optimization problems requires a large number of The computing power is used to solve nonlinear optimization problems, which leads to the technical problem of low efficiency of road decision-making.
  • the first aspect of the present application provides a road decision-making method, including:
  • the drivable road includes at least one lane, and each lane includes a plurality of waypoints connected in sequence;
  • the action at the current location is determined in real time according to the waypoint value of the next waypoint corresponding to the current location of the unmanned vehicle, and the current road decision result for driving to the destination is obtained.
  • the calculating the global cost from the target waypoint to the destination to obtain the waypoint value of the target waypoint includes:
  • the obtaining the global cost from the target waypoint to the destination based on the travel time of the unmanned vehicle from the target waypoint to the destination includes:
  • the upper The calculation process of the waypoint value of a waypoint is:
  • the waypoint value of the target waypoint On the basis of the waypoint value of the target waypoint, superimpose the short-term cost of transferring the previous waypoint corresponding to the target waypoint to the target waypoint, and calculate the target waypoint in combination with the state transition probability The waypoint value of the previous waypoint corresponding to the point.
  • a plurality of target waypoints are set between the current position of the unmanned vehicle and the destination.
  • the method also includes:
  • the updated waypoint value of the special waypoint iteratively update the waypoint values of the waypoints between the special waypoint and the current position of the unmanned vehicle in reverse iteration, and return the real-time information based on the unmanned vehicle
  • the traffic information includes static traffic participants
  • the acquisition of special waypoints affected by traffic information when the unmanned vehicle is traveling according to the current road decision-making result includes:
  • a special waypoint affected by the static traffic participant is determined according to the position of the static traffic participant.
  • the method also includes:
  • the static traffic participant When the special waypoint affected by the static traffic participant is a waypoint that will be reached in the future according to the current road decision result, and the adjacent lane of the lane where the special waypoint is located is impassable, the static traffic participant will be subject to the static waypoint.
  • the lane dividing line between the lane where the special waypoint affected by the traffic participant and the adjacent lane is divided into several waypoints connected in sequence;
  • the traffic information includes dynamic traffic participants
  • the acquisition of special waypoints affected by traffic information when the unmanned vehicle is traveling according to the current road decision-making result includes:
  • a special waypoint affected by the target dynamic traffic participant is determined according to the traveling speed of the target dynamic traffic participant and the traveling speed of the unmanned vehicle.
  • the determining the target dynamic traffic participant from the dynamic traffic participants includes:
  • a target traffic participant is determined from among the potential target dynamic traffic participants based on the confidence value.
  • the special waypoint is a waypoint that will be reached in the future according to the current road decision result
  • the next waypoint of the special waypoint is determined according to the current road decision result, and the special waypoint is updated.
  • the short-term cost of going to the next waypoint including:
  • the special waypoint is a waypoint that will be reached in the future according to the current road decision result, determine the next waypoint of the special waypoint according to the current road decision result, and determine the special waypoint to the next waypoint point travel distance;
  • the state transition probability includes a lane change success rate
  • the method further includes:
  • the updating the lane change success rate when transferring between waypoints within the preset range of the unmanned vehicle according to the traffic information includes:
  • the remaining waypoints within are other waypoints within the preset range of the unmanned vehicle except the current waypoint where the unmanned vehicle is located.
  • the calculation process of the current yield probability of the rear side vehicle of the unmanned vehicle is:
  • the second aspect of the application provides a road decision system, including:
  • a division module configured to divide the drivable road between the current position of the unmanned vehicle and the destination into several waypoints, the drivable road includes at least one lane, and each lane includes a plurality of waypoints connected in sequence;
  • the first calculation module is used to determine the target waypoint among the waypoints between the current position of the unmanned vehicle and the destination, and calculate the global cost from the target waypoint to the destination, and obtain the waypoint value of the target waypoint;
  • the second calculation module is used to iteratively calculate the waypoint value of the waypoint between the target waypoint and the current position of the unmanned vehicle according to the waypoint value of the target waypoint;
  • the decision-making module is used to determine the action at the current location according to the waypoint value of the next waypoint corresponding to the current location of the unmanned vehicle in real time, and obtain the current road decision-making result for driving to the destination.
  • the third aspect of the present application provides a road decision-making device, where the device includes a processor and a memory;
  • the memory is used to store program codes and transmit the program codes to the processor
  • the processor is configured to execute any one of the road decision-making methods described in the first aspect according to instructions in the program code.
  • the fourth aspect of the present application provides a computer-readable storage medium, the computer-readable storage medium is used to store program code, and when the program code is executed by a processor, any one of the road decision-making methods described in the first aspect is implemented. .
  • the present application has the following advantages:
  • the application provides a road decision-making method, including: dividing the drivable road between the current position of the unmanned vehicle and the destination into several waypoints, the drivable road includes at least one lane, each lane includes multiple Connected waypoints; determine the target waypoint among the waypoints between the current position of the unmanned vehicle and the destination, and calculate the global cost from the target waypoint to the destination to obtain the waypoint value of the target waypoint; according to the target waypoint Iteratively calculate the waypoint value of the waypoint between the target waypoint and the current position of the unmanned vehicle; determine the waypoint value at the current position according to the waypoint value of the next waypoint corresponding to the current position of the unmanned vehicle in real time.
  • the action is to get the current road decision-making result of driving to the destination, and the action is to turn left, keep the lane or turn right.
  • the drivable road between the unmanned vehicle and the destination is divided into waypoints, and by calculating the waypoint value of each waypoint, the unmanned vehicle can be used at each waypoint according to the waypoint of the next waypoint Values are used to make road decisions, which simplifies the complex road decision-making optimization problem, and calculates the waypoint value in two stages.
  • the first stage calculates the global cost from the target waypoint to the destination, and obtains the waypoint value of the target waypoint.
  • the waypoint value of the waypoint between the target waypoint and the current position of the unmanned vehicle is calculated in reverse iteration according to the waypoint value of the target waypoint, which improves the calculation speed of the waypoint value, thereby improving the efficiency of road decision-making.
  • FIG. 1 is a schematic flow chart of a road decision-making method provided in an embodiment of the present application
  • FIG. 2 is a distribution diagram of pivot points provided by the embodiment of the present application.
  • FIG. 3 is a waypoint distribution diagram provided in the embodiment of the present application.
  • FIG. 4 is a distribution diagram of the waypoint values of each waypoint in FIG. 3 obtained through static traffic information calculation provided by the embodiment of the present application;
  • Figure 5 is a traffic scene with static traffic participants provided by the embodiment of the present application.
  • Fig. 6 is the updated waypoint value distribution diagram of each waypoint value in Fig. 5 provided by the embodiment of the present application;
  • Fig. 7 is a traffic scene with dynamic traffic participants provided by the embodiment of the present application.
  • FIG. 8 is a distribution diagram of waypoint values before updating in the case of dynamic traffic participants provided by the embodiment of the present application.
  • FIG. 9 is a distribution diagram of the updated waypoint values in FIG. 8 provided by the embodiment of the present application.
  • FIG. 10 is a distribution diagram of waypoint values in a special traffic scenario provided by the embodiment of the present application.
  • FIG. 11 is a schematic structural diagram of a road decision-making system provided by an embodiment of the present application.
  • This application provides a road decision-making method, system, equipment, and medium, which are used to improve the prior art by adopting model predictive control methods to obtain optimal road decisions.
  • Obtaining optimal road decisions by solving complex optimization problems requires a large number of The computing power is used to solve nonlinear optimization problems, which leads to the technical problem of low efficiency of road decision-making.
  • the embodiment of the present application provides a road decision-making method, including:
  • Step 101 Divide the drivable road between the current position of the unmanned vehicle and the destination into several waypoints, the drivable road includes at least one lane, and each lane includes multiple waypoints connected in sequence.
  • the drivable road between the current position of the unmanned vehicle and the destination is divided into several waypoints.
  • the drivable road includes at least one lane, and each lane includes multiple waypoints connected in sequence.
  • the waypoints of each lane are evenly distributed.
  • Unmanned vehicles make road decisions at each waypoint to determine whether to change lanes and how to change lanes.
  • Step 102 Determine the target waypoint among the waypoints between the current position of the unmanned vehicle and the destination, and calculate the global cost from the target waypoint to the destination, and obtain the waypoint value of the target waypoint.
  • At least one target waypoint can be determined at a waypoint between the current position of the unmanned vehicle and the destination.
  • intersection including a crossroad, a T-junction, etc.
  • it can be The intersection of the drivable road between the current position of the unmanned vehicle and the destination is divided into a hub center, and the connection points between each road and each hub center are divided into a hub point, where the connection point includes the entry point of the hub center and For the exit point, please refer to Figure 2 for details.
  • the connection between the pivot point and the pivot point is the movement mode of the unmanned vehicle between the pivot points.
  • hub center is divided, you can choose the waypoint corresponding to the entry point of a certain hub center (it can be the hub center closest to the unmanned vehicle, such as hub center 1 in Figure 2) as the target waypoint, or you can choose multiple
  • the waypoint corresponding to the entry point of each hub center is used as the target waypoint. It is understandable that other waypoints may also be selected as the target waypoint, which is not specifically limited here.
  • the global cost from the target waypoint to the destination is calculated to obtain the waypoint value of the target waypoint.
  • the specific calculation process of the waypoint value of the target waypoint can be as follows:
  • the global map of the unmanned vehicle to the destination can be obtained, and then the global map can be analyzed by a graph search algorithm (such as the A-star algorithm) to obtain the shortest path from the target waypoint to the destination; then, based on The shortest path from the target waypoint to the destination and the preset travel speed calculate the travel time of the unmanned vehicle from the target waypoint to the destination, where the preset travel speed can be the speed limit value of the lane; finally, based on the The global cost from the target waypoint to the destination is obtained from the travel time from the target waypoint to the destination, and the global cost from the target waypoint to the destination is taken as the waypoint value of the target waypoint.
  • a graph search algorithm such as the A-star algorithm
  • the global map can be converted into a search map composed of hub points; when the target waypoint is not the entry point of the hub center, the global map can be converted into is a search graph composed of waypoints, and then the search graph is analyzed by a graph search algorithm to obtain the shortest path from the target waypoint to the destination.
  • the travel time of the unmanned vehicle from the target waypoint to the destination can be taken as the global cost from the target waypoint to the destination.
  • the target waypoint is the entry point of a certain hub center, at this time, there are multiple target waypoints.
  • the unmanned vehicle After calculating the global cost from each target waypoint to the destination, the unmanned vehicle can pass the global cost Determine which target waypoint enters the hub center to reach the destination the fastest.
  • the global cost from the target waypoint to the destination can be calculated according to the target information from the target waypoint to the destination and the travel time of the unmanned vehicle from the target waypoint to the destination.
  • the global cost from the target waypoint to the destination can be determined by multiple factors, for example, the distance between the target waypoint and the destination, the number of traffic lights between the target waypoint and the destination, whether there is a toll booth, etc. Therefore, the global cost can be calculated considering traffic light number information or toll station information etc. on the basis of travel time.
  • the target information in the embodiment of the present application includes traffic light number information or toll booth information, and the target information may also include other information related to driving needs.
  • the global cost can be obtained by linearly combining the travel time and the target information distribution weight.
  • the specific weight distribution can be set according to the actual situation, and is not specifically limited here.
  • the global cost of the target waypoint depends on the position of the destination input by the user, and when the user does not update the destination, the global cost of the target waypoint is fixed.
  • Step 103 iteratively calculating the waypoint values of the waypoints between the target waypoint and the current position of the unmanned vehicle according to the waypoint values of the target waypoint.
  • the waypoint selection process is modeled as a Markov decision model, and the waypoint is a state that an unmanned vehicle can be in.
  • the calculation process of the waypoint value of the previous waypoint corresponding to the target waypoint can be:
  • the step cost function C calculates the short-term cost when transferring between waypoints, and determines the state transition probability when transferring between waypoints through the transfer model T.
  • the last waypoint corresponding to the target waypoint can be calculated according to the distance between the last waypoint corresponding to the target waypoint and the target waypoint and the speed limit value or the average historical driving speed of the lane where the target waypoint is located. The travel time to change lanes, keep lanes, or turn right to the target waypoint.
  • the previous waypoint corresponding to the target waypoint includes the previous waypoint of the left lane, the previous waypoint of this lane and the last waypoint of the right lane ; If the target waypoint is located in the left lane of the three lanes, the last waypoint corresponding to the target waypoint includes the last waypoint of this lane and the last waypoint of the middle lane.
  • the travel time for transferring the previous waypoint corresponding to the target waypoint to the target waypoint may be directly used as the short-term cost for transferring the previous waypoint corresponding to the target waypoint to the target waypoint.
  • other losses may be considered on the basis of the travel time from the previous waypoint corresponding to the target waypoint calculated above to the target waypoint.
  • the previous waypoint corresponding to the target waypoint is superimposed on the basis of the waypoint value of the target waypoint and transferred to the target
  • the short-term cost of the waypoint is combined with the state transition probability to calculate the waypoint value of the previous waypoint corresponding to the target waypoint.
  • One waypoint is S3, the last waypoint in the adjacent lane (ie the middle lane) is S4, the target waypoint S1 is in the previous waypoint S4 of the lane (ie the middle lane), and the next waypoint in the adjacent lane (the left lane) is S4. and the right lane) are S3 and S5, the last waypoint of the target waypoint S2 is S5 in the lane (that is, the right lane), and the last waypoint in the adjacent lane (that is, the middle lane) for S4.
  • the actions that the unmanned vehicle can perform at the waypoint S4 include turning left, keeping the lane and turning right.
  • the executable actions of the unmanned vehicle at the waypoint S3 include keeping the lane and changing the right lane.
  • the executable actions at waypoint S5 include turning left and keeping the lane.
  • the waypoint values of waypoint S3 and waypoint S5 are shown in Figure 4.
  • V(s) min a ⁇ A ⁇ T [C(s,a,s')+V(s')];
  • V(s) is the waypoint value of waypoint s
  • C(s,a,s′) is the short-term cost of the unmanned vehicle to perform action a from waypoint s to waypoint s′
  • V(s′) is the waypoint value of waypoint s′
  • A is the executable action set of unmanned vehicle at waypoint s
  • ⁇ T ( ⁇ ) is the expected value function based on transfer model T.
  • the last waypoint corresponding to the target waypoint is used as the target waypoint, return to step S1031, and calculate the value of the last waypoint corresponding to the new target waypoint.
  • Waypoint values until the last waypoint corresponding to the target waypoint is the current position of the unmanned vehicle, and the waypoint values of all waypoints between the target waypoint and the current position of the unmanned vehicle are obtained.
  • the waypoints S3, S4, and S5 are used as new target waypoints. At this time, the waypoint S3 needs to be calculated.
  • the embodiment of this application is divided into two parts to calculate the waypoint value, one part is to obtain the global cost of the target waypoint through the global search of the target waypoint, and the other part is based on the value of the unmanned vehicle.
  • the short-term cost and state transition probability paid for transferring between different waypoints the global cost of the target waypoint is superimposed on the waypoint between the target waypoint and the current position of the unmanned vehicle, which reduces the amount of calculation and makes no Humans and vehicles can obtain optimal road decision-making results with a small amount of calculations, which improves the efficiency of road decision-making.
  • Step 104 Determine the action at the current location according to the waypoint value of the next waypoint corresponding to the current location of the unmanned vehicle in real time, and obtain the current road decision result for the destination.
  • the unmanned vehicle When the unmanned vehicle reaches the waypoint S1, it decides whether to change lanes and how to change lanes according to the minimum value of the waypoint S1 in the current lane and the next waypoint value of the adjacent lane of the current lane, so as to obtain the waypoint For the road decision result of S4, repeat the above steps to make a road decision, so as to drive to the destination.
  • the current distance between the target waypoint and the unmanned vehicle is calculated according to the waypoint value of the target waypoint.
  • the road decision is made in real time according to the waypoint value of the next waypoint corresponding to the current position of the unmanned vehicle; when the unmanned vehicle travels to the target waypoint according to the road decision result , the waypoint value of each waypoint between the destination and the current position of the unmanned vehicle (that is, the target waypoint) can be calculated according to the waypoint value of the destination, and then the value of the next waypoint corresponding to the target waypoint of the unmanned vehicle can be calculated in real time
  • the waypoint value of the waypoint is used to make road decisions to drive to the destination.
  • the waypoint value of the destination can be set to 0 or other relatively small values.
  • the calculation process of the waypoint values between waypoints is similar.
  • multiple target waypoints can be set between the current position of the unmanned vehicle and the destination at one time, and each target waypoint is separated by a certain distance along the driving direction of the unmanned vehicle .
  • the target waypoint that the unmanned vehicle arrives at first can be used as the first target waypoint (that is, the target waypoint closest to the unmanned vehicle), and the second arrived at The target waypoint is used as the second target waypoint (that is, the second closest target waypoint from the unmanned vehicle), and so on.
  • FIG. 2 Take Figure 2 as an example, assuming that the destination is a certain position in front of hub center 1, the unmanned vehicle is currently located behind hub center 2, and there are hub center 1 and hub center 2 between the destination and the current position of the unmanned vehicle, assuming that hub center 1 is selected.
  • the entry point of center 1 and hub center 2 is the target waypoint, which can be determined according to the distance between the entry point of hub center 1 and hub center 2 and the current position of the unmanned vehicle.
  • the waypoint corresponding to the entry point of hub center 2 is The first target waypoint
  • the waypoint corresponding to the entry point of the hub center 1 is the second target waypoint
  • the waypoint value of the entry point calculates the waypoint value of each waypoint between the entry point of the hub center 2 and the current position of the unmanned vehicle, and then in real time according to the waypoint value of the next waypoint corresponding to the current position of the unmanned vehicle Make road decisions; then, when the unmanned vehicle travels to a certain entry point of the hub center 2, the distance between the entry point of the hub center 1 and the entry point of the hub center 2 can be calculated according to the waypoint value of the entry point of the hub center 1
  • the waypoint value of the waypoint and then make road decisions based on the waypoint value, so as to drive to the destination.
  • multiple target waypoints can be selected to calculate the waypoint value of each waypoint in stages, and the total calculation amount is allocated to the calculation process of each stage, thereby increasing the calculation speed, and then Improve decision-making efficiency.
  • the unmanned vehicle takes each target waypoint as the destination of each stage, thereby gradually driving through each target waypoint, and finally reaches the destination.
  • the drivable road between the unmanned vehicle and the destination is divided into waypoints, and by calculating the waypoint value of each waypoint, the unmanned vehicle can be used at each waypoint according to the value of the next waypoint.
  • the waypoint value is used for road decision-making, which simplifies the complex road decision-making optimization problem, and calculates the waypoint value in two stages.
  • the first stage calculates the global cost from the target waypoint to the destination, and obtains the waypoint value of the target waypoint
  • the waypoint value of the waypoint between the target waypoint and the current position of the unmanned vehicle is iteratively calculated in reverse according to the waypoint value of the target waypoint, which improves the calculation speed of the waypoint value, thereby improving the road decision-making Efficiency, improving the existing technology using model predictive control method to obtain the optimal road decision, by solving complex optimization problems to obtain the optimal road decision, which requires a lot of computing power to solve the nonlinear optimization problem, resulting in low efficiency of road decision-making technical problems.
  • Step 201 Divide the drivable road between the current position of the unmanned vehicle and the destination into several waypoints, the drivable road includes at least one lane, and each lane includes multiple waypoints connected in sequence.
  • Step 202 Determine the target waypoint among the waypoints between the current position of the unmanned vehicle and the destination, and calculate the global cost from the target waypoint to the destination, and obtain the waypoint value of the target waypoint.
  • Step 203 iteratively calculating the waypoint values of the waypoints between the target waypoint and the current position of the unmanned vehicle according to the waypoint values of the target waypoint.
  • Step 204 Determine the action at the current location according to the waypoint value of the next waypoint corresponding to the current location of the unmanned vehicle in real time, and obtain the current road decision result for the destination.
  • steps 201 to 204 are consistent with the specific content of steps 101 to 104 described above, and will not be repeated here.
  • the above steps are to obtain waypoint values and make road decisions based on static traffic information, but the traffic environment in which unmanned vehicles are actually driving is dynamic and changes with time, and there are many other traffic participants. Participants will dynamically affect the one-step cost function and the transfer model of the unmanned vehicle, and finally affect the waypoint value of each waypoint. Therefore, during the driving process of unmanned vehicles, it is necessary to update the waypoint values according to the traffic information, and then update the road decision results.
  • the road decision-making method in the embodiment of the present application also includes:
  • Step 205 update the road decision result according to the traffic information.
  • the traffic information can be obtained in real time through the sensors on the unmanned vehicle or the Internet of Vehicles.
  • the traffic information includes static traffic participants (vehicles parked on the side of the road, traffic cones, etc.)
  • the location of the static traffic participants is determined according to the location of the static traffic participants.
  • special waypoints When a static traffic participant is at a certain waypoint, the waypoint is a special waypoint. Please refer to Figure 5.
  • an unmanned vehicle In a traffic scene, an unmanned vehicle is driving on the right lane and finds a traffic cone 30 meters ahead.
  • the right lane is blocked, and unmanned vehicles can predict that in the future, unmanned vehicles will not be able to drive to waypoint S2 by keeping the lane at waypoint S3, cannot drive to waypoint S2 through the right transition lane at waypoint S4, and cannot drive to waypoint S2 at waypoint S3.
  • S2 travels to waypoint S1 by keeping the lane, and cannot turn left at waypoint S2 to waypoint S0. That is, according to the position of the traffic cone, it can be determined that the special waypoint that will be affected by the traffic cone in the future is the waypoint S2.
  • the unmanned vehicle can predict that it will not be able to drive to the waypoint S2 by keeping the lane in the future. Point S1, so that the affected special waypoint can be determined as waypoint S2.
  • the target dynamic traffic participant is determined from the dynamic traffic participants; according to the driving speed of the target dynamic traffic participant and the driving speed of the unmanned vehicle Speed determination for special waypoints influenced by target dynamic traffic participants.
  • a target traffic participant is determined from potential target dynamic traffic participants based on the confidence value.
  • the behavior of dynamic traffic participants is uncertain.
  • it is necessary to determine which target traffic participants have the dynamic influence on the waypoint value For example, if the vehicle in front of the unmanned vehicle only starts to accelerate after driving slowly for 1 second, then the vehicle in front has little influence on the waypoint value, and the influence of the vehicle in front can be ignored. If the vehicle in front of the unmanned vehicle travels slowly After a period of time, it is necessary to consider the dynamic impact of the vehicle in front on the waypoint value.
  • a prior value can be configured for the potential target dynamic traffic participant, and whether the driving speed of the potential target dynamic traffic participant is less than the limit of the lane where the potential target dynamic traffic participant is located can be obtained.
  • the judgment result can be mapped to a numerical value through a mapping function. For example, the judgment result that the driving speed of the potential target dynamic traffic participant is less than the speed limit value of the lane can be mapped to a value 1, and the potential target dynamic The judgment result that the traffic participant's driving speed is greater than or equal to the speed limit value of the lane is mapped to a value of 0; and to get the confidence value of potential target dynamic traffic participants.
  • the potential target dynamic traffic participant When the confidence value of a potential target dynamic traffic participant is greater than the preset confidence threshold for a period of time, the potential target dynamic traffic participant is taken as the target dynamic traffic participant, which can avoid potential targets that suddenly accelerate or decelerate The dynamic traffic participant serves as the target dynamic traffic participant.
  • the special waypoint that will be affected by the target dynamic traffic participant in the future is determined according to the driving speed of the target dynamic traffic participant and the driving speed of the unmanned vehicle.
  • an unmanned vehicle (car1) is traveling at a constant speed v 1
  • the vehicle in front of the unmanned vehicle (car2) is traveling at a constant speed v 2 , where v 1 >v 2 , assuming that according to The current decision result of the unmanned vehicle based on the waypoint value calculated by the static traffic information is to keep the lane, that is, the right lane is the best lane at present.
  • the waypoint values of these waypoints do not consider the dynamic influence of the slow-moving vehicle in front.
  • v 1 >v 2 in a future area (the area is estimated by the speed difference between the unmanned vehicle and the vehicle in front), that is, In the shaded area in 7, the unmanned vehicle will be close to the vehicle in front, so that the unmanned vehicle will be affected by the slow-moving vehicle in the shadow area.
  • the short-term cost will increase, which will affect the waypoint value of waypoint S2, that is, waypoint S2 is a special waypoint that will be affected by the vehicle in front in the future.
  • the special waypoint is a waypoint affected by static traffic participants, as shown in Figure 5, the special waypoint is waypoint S2, and the next waypoint of waypoint S2 can be determined according to the current road decision result (lane keeping) is the waypoint S1, because there is a traffic cone at the waypoint S2, so that the unmanned vehicle cannot reach the waypoint S1 from the waypoint S2, the short-term cost C from the special waypoint S2 to the waypoint S1 can be updated (S2, keep the lane, S1 ) is a larger value (such as 50, 100, etc.), and the specific value can be set according to the actual situation.
  • a larger value such as 50, 100, etc.
  • the update process of the short-term cost from the special waypoint to the corresponding next waypoint is:
  • the special waypoint is a waypoint that will be reached in the future according to the current road decision result
  • the next waypoint of the special waypoint is determined according to the current road decision result, and the driving distance from the special waypoint to the next waypoint is determined;
  • the next waypoint of special waypoint S2 can be determined as waypoint S1, and according to the travel distance d between special waypoint S2 and waypoint S1 and the unmanned vehicle
  • the driving speed v 2 of the vehicle ahead can calculate the updated short-term cost s/v 2 from the special waypoint S2 to the waypoint S1.
  • the embodiment of this application further considers that the dynamic influence of the target dynamic traffic participant will last for a certain period of time, so , and finally the updated short-term cost C(s,a,s') of executing action a from a special waypoint s to the next waypoint s' can be expressed as:
  • ⁇ is a truncation parameter, which is used to determine the duration of the dynamic influence of the target dynamic traffic participant
  • d s'-s is the driving distance from the special waypoint s to the corresponding next waypoint s'
  • v is the target dynamic traffic Participant's driving speed.
  • the waypoint value of a waypoint is calculated from the waypoint value of the next waypoint corresponding to the waypoint, the short-term cost of transferring between waypoints and the state transition probability. After updating the short-term cost, the corresponding The waypoint value of will also be updated. It can be understood that if the state transition probability is updated, the corresponding waypoint value will also be updated.
  • the traffic cone is between waypoint S2 and waypoint S1, that is, the special waypoint S2 can turn left to waypoint S0
  • the short-term cost from waypoint S4 to special waypoint S2 also needs to be updated correspondingly.
  • the updating process of the waypoint value of waypoint S4 is similar to that of waypoint S3 and will not be repeated here.
  • the waypoint values of waypoint S3, waypoint S4 and the current position of the unmanned vehicle are updated iteratively in reverse. It should be noted that the short-term cost corresponding to the waypoints between waypoints S3 and S4 to the current position of the unmanned vehicle remains unchanged.
  • the updated waypoint value obtained is shown in Figure 6.
  • the unmanned vehicle will turn left to the left lane at the current waypoint , and go beyond the traffic cone.
  • the updated waypoint value obtained is shown in Figure 9. According to the updated waypoint value in Figure 9, it can be known that the unmanned vehicle will turn left to the left lane and overtake the front local.
  • the calculation of the waypoint value of each waypoint does not consider the time, that is, the influence of the dynamic traffic environment is not considered.
  • the unmanned vehicle needs to spend a huge amount of time to advance from the current waypoint to the next waypoint ahead, that is, the unmanned vehicle is between each waypoint.
  • the short-term cost paid for the transfer is closely related to the traffic environment.
  • the short-term cost will be updated according to the traffic information of each frame, which is dynamically changed.
  • the waypoint value is also dynamically changed.
  • the update formula of the waypoint value of each waypoint between the special waypoint and the current position of the unmanned vehicle can be expressed as:
  • V(s) is the updated waypoint value of waypoint s
  • C t (s,a,s′) is the short-term time for the unmanned vehicle to perform action a from waypoint s to reach waypoint s′ at current time t.
  • the cost, V(s′) is the waypoint value of the waypoint s′
  • A is the executable action set of the unmanned vehicle at the waypoint s, is based on the time-varying transfer model T t and the set of traffic participants at the current time t
  • the expected value function of is based on the time-varying transfer model T t and the set of traffic participants at the current time t.
  • the transfer model becomes time-dependent.
  • the current waypoint that is, the current state
  • the state reached by the unmanned vehicle to select an executable action is uncertain
  • the traffic density of the target lane change lane is close to its capacity, or the rear vehicle of the target lane change lane is approaching rapidly, even if the unmanned vehicle makes a lane change action, it may not be able to successfully change lanes to the target Change lanes. Therefore, it is necessary to dynamically update the success rate of lane change between waypoints by observing the traffic information around the unmanned vehicle.
  • the action that can be performed is determined by the lane where the unmanned vehicle is located. For example, the unmanned vehicle is in the rightmost lane, and there is no drivable road on the right side of the unmanned vehicle. At this time, turning right is an unexecutable action. Turning left is an action that can be performed.
  • the current distance between the vehicle behind the unmanned vehicle and the unmanned vehicle and the current yield probability of the vehicle behind the unmanned vehicle are obtained according to the traffic information, and the lane change success rate of the unmanned vehicle at the current waypoint is updated.
  • P(succ. t 1
  • d t ) is used to control the success rate of lane change according to the current distance between the unmanned vehicle and the rear side vehicle
  • P(succ. t 1
  • y t ) is used to control the lane change success rate according to the rear side vehicle
  • the cooperation of the vehicle is used to control the success rate of lane change
  • ⁇ is a proportional symbol.
  • P 0 is the lane change success rate at the current waypoint calculated in the static traffic environment, that is, the lane change success rate before the current waypoint is updated;
  • d t ) P 0 .
  • P(succ. t 1
  • y t ) is the current yield probability of the vehicle behind the unmanned vehicle
  • P(succ. t-1 1
  • y t-1 ) is the Yield probability at a moment
  • ⁇ is the update rate
  • at is the current acceleration of the rear side vehicle
  • the initial yield probability of the rear side vehicle is obtained by initialization, the initial yield probability of different rear side vehicles can be the same initial value, and the yield probability of the rear side vehicle can be updated according to the reaction of the rear side vehicle during driving .
  • the traffic density ⁇ t of the target lane change lane is updated according to
  • the lane change success rate of the remaining waypoints within the preset range of the unmanned vehicle, the target lane change lane is the lane after the lane change, which can be expressed as:
  • P(succ. t 1
  • ⁇ t ) is the lane change success rate of the remaining waypoints within the preset range of the unmanned vehicle under the traffic density at time t
  • ⁇ is the attenuation factor
  • ⁇ t is the target lane change
  • ⁇ is the traffic capacity of the target lane change lane
  • P max is the lane change success rate threshold.
  • the method in the embodiment of the present application also includes :
  • the lane separation line between the left lane and the right lane (that is, the solid line in Figure 10) can be divided into several sequentially connected waypoints, and then through the above steps
  • the waypoint value update formula in S2053 calculates the waypoint value of the waypoint on the lane dividing line, when the waypoint value of the waypoint on the dividing line is less than the updated waypoint value of the right lane affected by the traffic cone and the left lane
  • the unmanned vehicle can change lanes to the waypoint on the lane segmentation with less cost to surpass the traffic cone.
  • the waypoint value of the waypoint on the lane segmentation is shown in Figure 10.
  • the unmanned vehicle keeps going straight for a period of time Then, change lanes onto the lane divider to pass the traffic cone.
  • the embodiment of the present application considers that if the method of model predictive control is used to obtain the optimal road decision, it is necessary to obtain the optimal road decision by solving complex optimization problems, which requires a large amount of computing power to solve the nonlinear optimization problem, which relies heavily on The construction of environmental models is difficult to be effectively applied to the decision-making system of unmanned vehicles.
  • the embodiment of the present application is divided into two parts to solve the optimization problem, one part is to obtain the global cost of the target waypoint through the global search of the target waypoint, and the other part is to dynamically correct the transfer between different states by observing real-time traffic information
  • the short-term price paid and the success rate of lane change simplifies the optimization problem of high-dimensional multi-agents into the optimization problem of low-dimensional single agent, and the solution speed is faster.
  • the short-term cost of the road and the global cost are balanced, so that the unmanned vehicle can obtain the optimal road decision-making result with a small amount of calculation, thereby At the optimal time, follow the global navigation to actively change lanes, actively change lanes to super slow vehicles, actively change lanes to leave potential risk areas (such as construction areas, traffic accident areas, etc.), and actively change lanes to avoid priority vehicles (such as police cars, ambulances, etc.) )wait.
  • potential risk areas such as construction areas, traffic accident areas, etc.
  • priority vehicles such as police cars, ambulances, etc.
  • a road decision-making system provided by the embodiment of the present application, including:
  • the division module is used to divide the drivable road between the current position of the unmanned vehicle and the destination into several waypoints, the drivable road includes at least one lane, and each lane includes a plurality of waypoints connected in sequence;
  • the first calculation module is used to determine the target waypoint among the waypoints between the current position of the unmanned vehicle and the destination, and calculate the global cost from the target waypoint to the destination, and obtain the waypoint value of the target waypoint;
  • the second calculation module is used to iteratively calculate the waypoint value of the waypoint between the target waypoint and the current position of the unmanned vehicle according to the waypoint value of the target waypoint;
  • the decision-making module is used to determine the action at the current position according to the waypoint value of the next waypoint corresponding to the current position of the unmanned vehicle in real time, and obtain the current road decision-making result towards the destination.
  • the action is to turn left, keep the lane or Turn right.
  • the first calculation module is specifically used for:
  • the global cost from the target waypoint to the destination is taken as the waypoint value of the target waypoint.
  • the road decision-making system in the embodiment of the present application also includes: a waypoint value update module, used for:
  • the special waypoint is a waypoint that will be reached in the future according to the current road decision result, determine the next waypoint of the special waypoint according to the current road decision result, and update the short-term cost from the special waypoint to the next waypoint;
  • the waypoint value of each waypoint between the special waypoint and the current position of the unmanned vehicle is updated iteratively in reverse according to the updated waypoint value of the special waypoint, and a decision-making module is triggered.
  • the road decision system in the embodiment of the present application also includes: a third calculation module, used for:
  • the special waypoint affected by static traffic participants is a waypoint that will be reached in the future according to the current road decision result, and the adjacent lane of the lane where the special waypoint is located is impassable, the special waypoint that will be affected by static traffic participants
  • the lane dividing line between the lane where it is located and the adjacent lane is divided into several waypoints connected in sequence;
  • the state transition probability includes the lane change success rate
  • the road decision system in the embodiment of the present application also includes:
  • the lane change success rate update module is used to update the lane change success rate when transferring between waypoints within the preset range of the unmanned vehicle according to traffic information.
  • the lane change success rate update module is specifically used for:
  • the target lane change lane is the lane after the lane change
  • the remaining waypoints within the preset range of the unmanned vehicle are Other waypoints within the preset range of the unmanned vehicle except the current waypoint where the unmanned vehicle is located.
  • the drivable road between the unmanned vehicle and the destination is divided into waypoints, and by calculating the waypoint value of each waypoint, the unmanned vehicle can be used at each waypoint according to the value of the next waypoint.
  • the waypoint value is used for road decision-making, which simplifies the complex road decision-making optimization problem, and calculates the waypoint value in two stages.
  • the first stage calculates the global cost from the target waypoint to the destination, and obtains the waypoint value of the target waypoint
  • the waypoint value of the waypoint between the target waypoint and the current position of the unmanned vehicle is iteratively calculated in reverse according to the waypoint value of the target waypoint, which improves the calculation speed of the waypoint value, thereby improving the road decision-making Efficiency, improving the existing technology using model predictive control method to obtain the optimal road decision, by solving complex optimization problems to obtain the optimal road decision, which requires a lot of computing power to solve the nonlinear optimization problem, resulting in low efficiency of road decision-making technical problems.
  • the embodiment of the present application also provides a road decision-making device, which includes a processor and a memory;
  • the memory is used to store the program code and transmit the program code to the processor
  • the processor is configured to execute the road decision-making method in the aforementioned method embodiments according to the instructions in the program code.
  • the embodiment of the present application also provides a computer-readable storage medium, and the computer-readable storage medium is used for storing program codes.
  • the program codes are executed by a processor, the road decision-making method in the aforementioned method embodiments is implemented.
  • At least one (item) means one or more, and “multiple” means two or more.
  • “And/or” is used to describe the association relationship of associated objects, indicating that there can be three types of relationships, for example, “A and/or B” can mean: only A exists, only B exists, and A and B exist at the same time , where A and B can be singular or plural.
  • the character “/” generally indicates that the contextual objects are an “or” relationship.
  • At least one of the following” or similar expressions refer to any combination of these items, including any combination of single or plural items.
  • At least one item (piece) of a, b or c can mean: a, b, c, "a and b", “a and c", “b and c", or "a and b and c ", where a, b, c can be single or multiple.
  • the disclosed devices and methods may be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components can be combined or May be integrated into another system, or some features may be ignored, or not implemented.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.
  • the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the technical solution of the present application is essentially or part of the contribution to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions for executing all or part of the steps of the methods described in the various embodiments of the present application through a computer device (which may be a personal computer, a server, or a network device, etc.).
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (English full name: Read-Only Memory, English abbreviation: ROM), random access memory (English full name: Random Access Memory, English abbreviation: RAM), magnetic Various media that can store program codes such as discs or optical discs.

Landscapes

  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Human Computer Interaction (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Navigation (AREA)
  • Traffic Control Systems (AREA)

Abstract

一种道路决策方法、***、设备和介质，将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点；在无人车辆的当前位置到目的地之间的航点中确定目标航点，并计算目标航点到目的地的全局代价，得到目标航点的航点值；根据目标航点的航点值迭代计算目标航点到无人车辆的当前位置之间的航点的航点值；实时根据无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向目的地的当前道路决策结果，改善了现有技术采用模型预测控制的方法来获取最优道路决策，通过求解复杂的优化问题来得到最优道路决策，需要大量的运算能力来求解非线性优化问题，导致道路决策效率低的技术问题。

Description

一种道路决策方法、***、设备和介质
本申请要求于2021年09月29日提交中国专利局、申请号为CN202111155395.9、发明名称为“一种无人驾驶车辆变道决策方法、***、设备和介质”和2021年11月30日提交中国专利局、申请号为CN202111453193.2、发明名称为“一种道路决策方法、***、设备和介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及无人驾驶车辆技术领域，尤其涉及一种道路决策方法、***、设备和介质。
背景技术
无人车辆为集环境感知与认知、动态规划与决策、行为控制与执行等多项功能于一体的综合智能平台，无人车辆研究的核心问题包括环境感知、行为决策和运动控制。
道路决策是无人车辆决策技术的主要组成部分，现有技术通常采用模型预测控制的方法来获取最优道路决策，通过求解复杂的优化问题来得到最优道路决策，需要大量的运算能力来求解非线性优化问题，导致道路决策效率低，难以被有效应用到无人驾驶车辆的决策***中。
发明内容
本申请提供了一种道路决策方法、***、设备和介质，用于改善现有技术采用模型预测控制的方法来获取最优道路决策，通过求解复杂的优化问题来得到最优道路决策，需要大量的运算能力来求解非线性优化问题，导致道路决策效率低的技术问题。
有鉴于此，本申请第一方面提供了一种道路决策方法，包括：
将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，所述可行驶道路至少包括一条车道，每条车道包括多个依次相连的航点；
在所述无人车辆的当前位置到所述目的地之间的航点中确定目标航点，并计算所述目标航点到所述目的地的全局代价，得到所述目标航点的 航点值；
根据所述目标航点的航点值迭代计算所述目标航点到所述无人车辆的当前位置之间的航点的航点值；
实时根据所述无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向所述目的地的当前道路决策结果。
可选的，所述计算所述目标航点到所述目的地的全局代价，得到所述目标航点的航点值，包括：
通过图搜索算法获取所述目标航点到所述目的地的最短路径；
基于所述最短路径和预置行驶速度计算所述无人车辆从所述目标航点到所述目的地的行驶时间；
基于所述无人车辆从所述目标航点到所述目的地的行驶时间获取所述目标航点到所述目的地的全局代价；
将所述目标航点到所述目的地的全局代价作为所述目标航点的航点值。
可选的，所述基于所述无人车辆从所述目标航点到所述目的地的行驶时间获取所述目标航点到所述目的地的全局代价，包括：
将所述无人车辆从所述目标航点到所述目的地的行驶时间作为所述目标航点到所述目的地的全局代价；
或，根据所述目标航点到所述目的地的目标信息和所述无人车辆从所述目标航点到所述目的地的行驶时间计算所述目标航点到所述目的地的全局代价。
可选的，在根据所述目标航点的航点值迭代计算所述目标航点到所述无人车辆的当前位置之间的航点的航点值时，所述目标航点对应的上一个航点的航点值的计算过程为：
计算所述目标航点对应的上一个航点转移到所述目标航点的短期代价和状态转移概率；
在所述目标航点的航点值的基础上叠加所述目标航点对应的上一个航点转移到所述目标航点的所述短期代价，并结合所述状态转移概率计算所述目标航点对应的上一个航点的航点值。
可选的，所述无人车辆的当前位置到所述目的地之间设置有多个所述 目标航点。
可选的，所述方法还包括：
在所述无人车辆根据当前道路决策结果行驶时，获取受到交通信息影响的特殊航点；
当所述特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定所述特殊航点的下一个航点，并更新所述特殊航点到该下一个航点的短期代价；
基于所述特殊航点到该下一个航点更新后的短期代价更新所述特殊航点的航点值；
根据所述特殊航点更新后的航点值反向迭代更新所述特殊航点到所述无人车辆的当前位置之间的各航点的航点值，并返回所述实时根据所述无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向所述目的地的当前道路决策结果的步骤。
可选的，当所述交通信息包括静态交通参与者时；
所述在所述无人车辆根据当前道路决策结果行驶时，获取受到交通信息影响的特殊航点，包括：
在所述无人车辆根据当前道路决策结果行驶时，根据所述静态交通参与者的位置确定受到所述静态交通参与者的影响的特殊航点。
可选的，所述方法还包括：
当受到所述静态交通参与者影响的所述特殊航点为根据当前道路决策结果行驶未来会到达的航点，且所述特殊航点所在车道的相邻车道无法通行时，将受到所述静态交通参与者影响的所述特殊航点所在车道与该相邻车道之间的车道分隔线划分为若干个依次相连的航点；
根据所述车道分割线的相邻车道上的航点的航点值、各航点之间转移的短期代价和状态转移概率计算该车道分隔线上各航点的航点值，并返回所述实时根据所述无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向所述目的地的当前道路决策结果的步骤。
可选的，当所述交通信息包括动态交通参与者时；
所述在所述无人车辆根据当前道路决策结果行驶时，获取受到交通信息影响的特殊航点，包括：
在所述无人车辆根据当前道路决策结果行驶时，从所述动态交通参与者中确定目标动态交通参与者；
根据所述目标动态交通参与者的行驶速度和所述无人车辆的行驶速度确定受到所述目标动态交通参与者的影响的特殊航点。
可选的，所述从所述动态交通参与者中确定目标动态交通参与者，包括：
将位于所述无人车辆前方预置范围内的动态交通参与者作为潜在目标动态交通参与者；
判断所述潜在目标动态交通参与者的行驶速度是否小于所述潜在目标动态交通参与者所在车道的限速值，得到判断结果；
根据所述潜在目标动态交通参与者的值和所述判断结果计算所述潜在目标动态交通参与者的置信度值；
基于所述置信度值从所述潜在目标动态交通参与者中确定目标交通参与者。
可选的，所述当所述特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定所述特殊航点的下一个航点，并更新所述特殊航点到该下一个航点的短期代价，包括：
当所述特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定所述特殊航点的下一个航点，并确定所述特殊航点到该下一个航点的行驶距离；
根据所述行驶距离和所述目标交通参与者的行驶速度计算所述无人车辆从所述特殊航点到该下一个航点的短期代价，得到所述特殊航点到该下一个航点的更新后的短期代价。
可选的，所述状态转移概率包括变道成功率，所述方法还包括：
根据所述交通信息更新所述无人车辆预置范围内的航点之间转移时的变道成功率。
可选的，所述根据所述交通信息更新所述无人车辆预置范围内的航点之间转移时的变道成功率，包括：
根据所述交通信息获取所述无人车辆的后侧方车辆与所述无人车辆的当前距离和所述无人车辆的后侧方车辆的当前让步概率，更新所述无人车 辆在当前航点的变道成功率；
根据目标变道车道的交通密度更新在所述无人车辆预置范围内的剩余航点的变道成功率，所述目标变道车道为变道后的车道，所述无人车辆预置范围内的剩余航点为所述无人车辆预置范围内的除所述无人车辆所在的当前航点之外的其他航点。
可选的，所述无人车辆的后侧方车辆的当前让步概率的计算过程为：
根据所述无人车辆的后侧方车辆的当前加速度和该后侧方车辆在前一时刻的让步概率计算该后侧方车辆的当前让步概率，其中，该后侧方车辆的初始让步概率通过初始化得到。
本申请第二方面提供了一种道路决策***，包括：
划分模块，用于将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，所述可行驶道路至少包括一条车道，每条车道包括多个依次相连的航点；
第一计算模块，用于在所述无人车辆的当前位置到所述目的地之间的航点中确定目标航点，并计算所述目标航点到所述目的地的全局代价，得到所述目标航点的航点值；
第二计算模块，用于根据所述目标航点的航点值迭代计算所述目标航点到所述无人车辆的当前位置之间的航点的航点值；
决策模块，用于实时根据所述无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向所述目的地的当前道路决策结果。
本申请第三方面提供了一种道路决策设备，所述设备包括处理器以及存储器；
所述存储器用于存储程序代码，并将所述程序代码传输给所述处理器；
所述处理器用于根据所述程序代码中的指令执行第一方面任一种所述的道路决策方法。
本申请第四方面提供了一种计算机可读存储介质，所述计算机可读存储介质用于存储程序代码，所述程序代码被处理器执行时实现第一方面任一种所述的道路决策方法。
从以上技术方案可以看出，本申请具有以下优点：
本申请提供了一种道路决策方法，包括：将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，可行驶道路至少包括一条车道，每条车道包括多个依次相连的航点；在无人车辆的当前位置到目的地之间的航点中确定目标航点，并计算目标航点到目的地的全局代价，得到目标航点的航点值；根据目标航点的航点值迭代计算目标航点到无人车辆的当前位置之间的航点的航点值；实时根据无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向目的地的当前道路决策结果，动作为左转变道、保持车道或右转变道。
本申请中，将无人车辆到目的地之间的可行驶道路划分为航点，通过计算各航点的航点值，使得无人车辆在每个航点可以根据下一个航点的航点值进行道路决策，实现了将复杂的道路决策优化问题进行简化，并且分两阶段计算航点值，第一阶段计算目标航点到目的地的全局代价，得到目标航点的航点值，第二阶段根据目标航点的航点值反向迭代计算目标航点到无人车辆的当前位置之间的航点的航点值，提高了航点值的计算速度，从而提高了道路决策效率，改善了现有技术采用模型预测控制的方法来获取最优道路决策，通过求解复杂的优化问题来得到最优道路决策，需要大量的运算能力来求解非线性优化问题，导致道路决策效率低的技术问题。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其它的附图。
图1为本申请实施例提供的一种道路决策方法的一个流程示意图；
图2为本申请实施例提供的一种枢纽点分布图；
图3为本申请实施例提供的一种航点分布图；
图4为本申请实施例提供的通过静态交通信息计算得到的图3中各航点的航点值的分布图；
图5为本申请实施例提供的有静态交通参与者的一个交通场景；
图6为本申请实施例提供的图5中各航点值后更新后的航点值分布图；
图7为本申请实施例提供的有动态交通参与者的一个交通场景；
图8为本申请实施例提供的在有动态交通参与者情况下更新前的航点值分布图；
图9为本申请实施例提供的图8中各航点值更新后的航点值分布图；
图10为本申请实施例提供的一种特殊交通场景下的航点值分布图；
图11为本申请实施例提供的一种道路决策***的一个结构示意图。
具体实施方式
本申请提供了一种道路决策方法、***、设备和介质，用于改善现有技术采用模型预测控制的方法来获取最优道路决策，通过求解复杂的优化问题来得到最优道路决策，需要大量的运算能力来求解非线性优化问题，导致道路决策效率低的技术问题。
为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
为了便于理解，请参阅图1，本申请实施例提供了一种道路决策方法，包括：
步骤101、将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，可行驶道路至少包括一条车道，每条车道包括多个依次相连的航点。
在确定了目的地后，将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，可行驶道路至少包括一条车道，每条车道包括多个依次相连的航点，各车道的航点均匀分布。无人车辆在各个航点进行道路决策，确定是否进行变道，以及如何变道。
步骤102、在无人车辆的当前位置到目的地之间的航点中确定目标航点，并计算目标航点到目的地的全局代价，得到目标航点的航点值。
可以在无人车辆的当前位置到目的地之间的航点确定至少一个目标航 点，当无人车辆的当前位置到目的地之间存在交叉口(包括十字路口、丁字路口等)时，可以将无人车辆的当前位置到目的地之间的可行驶道路的交叉口划分为枢纽中心，将各道路与各枢纽中心的连接点划分为枢纽点，其中，连接点包括枢纽中心的入口点和出口点，具体可以参考图2，枢纽点与枢纽点之间的连线为无人车辆在枢纽点之间的移动方式，例如，无人车辆可以由当前枢纽点通过变道等动作到达另一个枢纽点，但无人车辆在枢纽中心不能进行变道，枢纽点与枢纽点之间的连接关系需考虑全局地图和交通规则。在划分得到枢纽中心后，可以选择某个枢纽中心(可以是距离无人车辆最近的枢纽中心，如图2中的枢纽中心1)的入口点对应的航点作为目标航点，也可以选择多个枢纽中心的入口点对应的航点作为目标航点。可以理解的是，也可以选择其他的航点作为目标航点，在此不做具体限定。
在确定目标航点后，计算目标航点到目的地的全局代价，得到目标航点的航点值。目标航点的航点值的具体计算过程可以为：
通过图搜索算法获取目标航点到目的地的最短路径；基于最短路径和预置行驶速度计算无人车辆从目标航点到目的地的行驶时间；基于无人车辆从目标航点到目的地的行驶时间获取目标航点到目的地的全局代价；将目标航点到目的地的全局代价作为目标航点的航点值。
具体的，可以获取无人车辆到目的地的全局地图，然后可以通过图搜索算法(例如A-star算法)对该全局地图进行分析，以获取目标航点到目的地的最短路径；然后，基于目标航点到目的地的最短路径和预置行驶速度计算无人车辆从目标航点到目的地的行驶时间，其中，该预置行驶速度可以为车道的限速值；最后，基于无人车辆从目标航点到目的地的行驶时间获取目标航点到目的地的全局代价，将目标航点到目的地的全局代价作为目标航点的航点值。
需要说明的是，当目标航点为枢纽中心的入口点时，可以将全局地图转换为一张由枢纽点构成的搜索图；当目标航点不是枢纽中心的入口点时，可以将全局地图转换为由航点构成的搜索图，进而通过图搜索算法对该搜索图进行分析，以获取目标航点到目的地的最短路径。
在一种实施例中，可以将无人车辆从目标航点到目的地的行驶时间作 为目标航点到目的地的全局代价。当目标航点为某枢纽中心的入口点时，此时，目标航点的数量存在多个的情况，在计算得到各目标航点到目的地的全局代价后，无人车辆通过该全局代价可以确定由哪个目标航点进入枢纽中心可以最快到达目的地。
在另一种实施例中，可以根据目标航点到目的地的目标信息和无人车辆从目标航点到目的地的行驶时间计算目标航点到目的地的全局代价。目标航点到目的地的全局代价可以由多个因素确定，例如，目标航点到目的地之间的距离、目标航点到目的地的之间的红绿灯数量、是否有收费站等。因此，可以在行驶时间的基础上考虑红绿灯数量信息或收费站信等来计算全局代价。具体的，综合考虑行驶时间和目标信息，本申请实施例中的目标信息包括红绿灯数量信息或收费站信息，目标信息还可以包括其他驾驶需求相关的信息。可以对行驶时间和目标信息分配权重进行线性组合来获取全局代价，具体的权重分配情况可以根据实际情况进行设置，在此不作具体限定。
可以理解的是，目标航点的全局代价取决于用户输入目的地的位置，在用户没有更新目的地时，目标航点的全局代价是固定的。
步骤103、根据目标航点的航点值迭代计算目标航点到无人车辆的当前位置之间的航点的航点值。
本申请实施例中，将航点的选择过程建模为马尔可夫决策模型，航点为无人车辆可处于的状态。马尔可夫决策模型可以表示为<S,A,T,C>，S为无人车辆的状态空间，A＝{左转变道，保持车道，右转变道}为无人车辆的动作集合，C为单步代价函数，用于计算无人车辆从一个状态转移到另一个状态所需付出的短期代价，例如，C(s,a,s')用于计算无人车辆从航点s执行动作a转移到航点s'所需付出的短期代价；T为转移模型，表示由行动引起的不确定性，例如，T(s,右转变道,s')表示无人车辆在航点执行右转变道转移到航点的变道成功率。
本申请实施例中，在根据目标航点的航点值迭代计算目标航点到无人车辆的当前位置之间的航点的航点值时，只考虑静态交通信息，假设无人车辆在一个静态和时间不变的交通环境中行驶，其中，无人车辆的当前位置到目的地之间仅有本车，没有其他的交通参与者。其中，目标航点对应 的上一个航点的航点值的计算过程可以为：
S1031、计算目标航点对应的上一个航点转移到目标航点的短期代价和状态转移概率。
在计算得到目标航点的航点值后，反向迭代计算目标航点对应的上一个航点的航点值，而无人车辆在航点之间转移时需要付出一定的代价，可以通过单步代价函数C计算各航点之间转移时的短期代价，通过转移模型T确定各航点之间转移时的状态转移概率。其中，可以根据目标航点对应的上一个航点与目标航点之间的距离和目标航点所在车道的限速值或历史行驶速度均值，计算出目标航点对应的上一个航点执行左转变道、保持车道或右转变道转移到目标航点的行驶时间。需要说明的是，目标航点对应的上一个航点存在多个航点的情况。假设目标航点位于三车道的中间车道，此时，该目标航点对应的上一个航点包括左侧车道的上一个航点、本车道的上一个航点以及右侧车道的上一个航点；若目标航点位于三车道的左侧车道，此时该目标航点对应的上一个航点包括本车道的上一个航点和中间车道的上一个航点。
在一种实施例中，可以直接将目标航点对应的上一个航点转移到目标航点的行驶时间作为目标航点对应的上一个航点转移到目标航点的短期代价。
在另一种实施例中，可以在上述计算得到的目标航点对应的上一个航点转移到目标航点的行驶时间的基础上考虑其他的损失，例如，不想让无人车辆行驶在最右车道或者不想让无人车辆进入公交车道等的用户喜好设置，这些用户喜好设置会产生一定的损失，因此，可以在目标航点对应的上一个航点转移到目标航点的行驶时间的基础上增加由用户喜好设置所产生的损失，以得到目标航点对应的上一个航点转移到目标航点的短期代价。
S1032、在目标航点的航点值的基础上叠加目标航点对应的上一个航点转移到目标航点的短期代价，并结合状态转移概率计算目标航点对应的上一个航点的航点值。
在反向迭代计算目标航点到无人车辆的当前位置之间的航点的航点值时，在目标航点的航点值的基础上叠加目标航点对应的上一个航点转移到目标航点的短期代价，并结合状态转移概率计算目标航点对应的上一个航 点的航点值。目标航点的上一个航点存在多个的情况，不同的航点转移到目标航点执行的动作不相同，如图3所示，目标航点S0在所在车道(即左侧车道)的上一个航点为S3，在相邻车道(即中间车道)的上一个航点为S4，目标航点S1在所在车道(即中间车道)的上一个航点S4，在相邻车道(左侧车道和右侧车道)的上一个航点为S3、S5，目标航点S2在所在车道(即右侧车道)的上一个航点为S5，在相邻车道(即中间车道)的上一个航点为S4。
假设目标航点S0、S1、S2的航点值分别为100、50、80，单步代价函数设置为C(s,a,s')＝1，即无人车辆在静态交通环境中只需支付1个单位的代价在各航点之间转移，变道成功率设置为20％，即无人车辆在各航点进行变道时，有20％的机会变道成功。
在计算目标航点对应的上一航点S4的航点值时，无人车辆在航点S4的可执行的动作包括左转变道、保持车道和右转变道。当在航点S4选择保持车道转移到目标航点S1时，对应的航点S4的航点值为V(S4) 保持车道＝(50+1)*100％＝51；
当在航点S4选择左转变道转移到目标航点S0时，由于变道成功率为20％，对应的航点S4的航点值为V(S4) 左转变道＝(100+1)*20％+(50+1)*80％＝71；
当在航点S4选择右转变道转移到航点目标航点S2时，由于变道成功率为20％，对应的航点S4的航点值为V(S4) 右转变道＝(80+1)*20％+(50+1)*80％＝57；
最终V(S4)＝min(V(S4) 保持车道,V(S4) 左转变道,V(S4) 右转变道)＝51，因此，最终航点S4的航点值为51。
无人车辆在航点S3的可执行动作包括保持车道和右转变道，航点S3的航点值为V(S3)＝min(V(S3) 保持车道,V(S3) 右转变道)，在航点S5的可执行动作包括左转变道和保持车道，航点S5的航点值为V(S5)＝min(V(S5) 保持车道,V(S5) 左转变道)，最终计算得到的航点S3和航点S5的航点值如图4所示。
在目标航点到无人车辆的当前位置之间的各航点的航点值的计算过程可以归纳为：
V(s)＝min a∈AΕ T[C(s,a,s′)+V(s′)]；
式中，V(s)为航点s的航点值，C(s,a,s′)为无人车辆从航点s执行动作 a到达航点s′的短期代价，V(s′)为航点s′的航点值，A为无人车辆在航点s的可执行动作集合，Ε T(·)为基于转移模型T的期望值函数。
在计算得到目标航点对应的上一个航点的航点值后，将目标航点对应的上一个航点作为目标航点，返回步骤S1031，计算新的目标航点对应的上一个航点的航点值，直至目标航点对应的上一个航点为无人车辆的当前位置，得到目标航点到无人车辆的当前位置之间的所有航点的航点值。请参考图4，在计算得到目标航点对应的上一个航点S3、S4、S5的航点值后，将航点S3、S4、S5作为新的目标航点，此时需要计算航点S3、S4、S5对应的上一个航点S6、S7、S8的航点值，在计算得到航点S6、S7、S8转移到对应的下一个航点的短期代价和状态转移概率后，在航点S3、S4、S5的航点值的基础上叠加对应的短期代价，并结合状态转移概率计算得到航点S6、S7、S8的航点值，然后将航点S6、S7、S8作为新的目标航点，计算航点S6、S7、S8对应的上一个航点的航点值，以此类推，迭代计算得到目标航点到无人车辆的当前位置之间的各航点的航点值。
若直接从目的地方向迭代计算目的地到无人车辆之间的各航点的航点值，来获取最优的道路决策，当无人车辆距离目的地较远时，该过程计算量很大，影响无人车辆的道路决策效率；而本申请实施例分两部分来计算航点值，一部分是通过对目标航点的全局搜索来获取目标航点的全局代价，另一部分是根据无人车辆在不同航点之间转移所付出的短期代价与状态转移概率，将目标航点的全局代价叠加到目标航点到无人车辆的当前位置之间的航点上，减少了计算量，使得无人车辆可以以少量的运算获取最优道路决策结果，提高了道路决策效率。
步骤104、实时根据无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向目的地的当前道路决策结果。
实时根据各航点当前的航点值进行道路决策，确定在当前位置的动作，是要左转变道，还是保持车道，还是右转变道。具体的，实时根据无人车辆当前位置所在的当前车道及当前车道的相邻车道的下一个航点的航点值，确定无人车辆在当前位置对应的下一个航点的航点值中的最小航点值，根据该最小航点值对应的航点所在的位置确定是否变道以及如何变道。请参考图4，假设无人车辆当前位于航点S4，根据航点S4对应的下一个航 点S0、S1、S2的航点值，可以确定航点S1的航点值最小，而航点S1位于航点S4正前方，即航点S1与航点S4位于同一车道，因此，无人车辆在航点S4选择保持车道直行到航点S1，即在航点S4的道路决策结果为保持车道。当无人车辆到达航点S1后，根据航点S1在当前车道及当前车道的相邻车道的下一个航点的航点值的最小值决策是否变道以及如何变道，从而得到在航点S4的道路决策结果，重复上述步骤进行道路决策，从而驶向目的地。
在一种实施例中，在无人车辆的当前位置到目的地之间的目标航点的数量为1个时，在根据目标航点的航点值计算得到目标航点到无人车辆的当前位置之间的各航点的航点值后，实时根据无人车辆在当前位置对应的下一个航点的航点值进行道路决策；在无人车辆根据道路决策结果行驶的该目标航点时，可以根据目的地的航点值计算目的地到无人车辆的当前位置(即目标航点)之间的各航点的航点值，然后实时根据无人车辆在目标航点对应的下一个航点的航点值进行道路决策，从而驶向目的地。其中，目的地的航点值可以设置为0或其他相对较小的数值，目的地到目标航点之间的航点的航点值的计算过程与目标航点到无人车辆的当前位置之间的航点的航点值的计算过程类似。
在另一种实施例中，为了进一步提高计算效率，可以一次性在无人车辆的当前位置到目的地之间设置多个目标航点，各目标航点沿无人车辆的行驶方向间隔一定距离。在通过步骤102计算得到各目标航点的航点值后，可以将无人车辆最先到达的目标航点作为第一目标航点(即距离无人车辆最近的目标航点)，第二到达的目标航点作为第二目标航点(即距离无人车辆第二近的目标航点)，以此类推。根据第一目标航点的航点值计算第一目标航点到无人车辆的当前位置之间的各航点的航点值，然后实时根据无人车辆在当前位置对应的下一个航点的航点值进行道路决策；当无人车辆根据道路决策结果行驶到第一目标航点后，根据第二目标航点的航点值计算第二目标航点到无人车辆的当前位置(即第一目标航点)之间的各航点的航点值，以此类推，当无人车辆到达最后的目标航点时，可以根据目的地的航点值计算目的地到无人车辆的当前位置(最后的目标航点)之间的各航点的航点值，然后实时根据最后的目标航点对应的下一个航点的航 点值进行道路决策，从而驶向目的地。
以图2为例，假设目的地为枢纽中心1前方某一位置，无人车辆当前位于枢纽中心2后方，目的地与无人车辆当前位置之间存在枢纽中心1和枢纽中心2，假设选择枢纽中心1和枢纽中心2的入口点为目标航点，根据枢纽中心1和枢纽中心2的入口点与无人车辆的当前位置之间的距离可以确定，枢纽中心2的入口点对应的航点为第一目标航点，枢纽中心1的入口点对应的航点为第二目标航点，在根据目标航点的航点值计算其他航点的航点值时，首先，可以根据枢纽中心2的入口点的航点值计算枢纽中心2的入口点到无人车辆的当前位置之间的各航点的航点值，然后实时根据无人车辆在当前位置对应的下一个航点的航点值进行道路决策；然后，在无人车辆行驶到枢纽中心2的某一个入口点时，可以根据枢纽中心1的入口点的航点值计算枢纽中心1的入口点到枢纽中心2的入口点之间的航点的航点值，然后进行道路决策；当无人车辆行驶到枢纽中心1的某个入口点时，可以根据目的地的航点值计算目的地到枢纽中心1的入口点之间的航点的航点值，再根据航点值进行道路决策，从而驶向目的地。在无人车辆距离目的地较远时，可以选择多个目标航点来分阶段计算各航点的航点值，将总的计算量分摊到各个阶段的计算过程中，从而提高计算速度，进而提高决策效率。通过设置多个目标航点，使得无人车辆将各目标航点作为各个阶段的目的地，由此逐步行驶经过各目标航点，最终到达目的地。
本申请实施例中，将无人车辆到目的地之间的可行驶道路划分为航点，通过计算各航点的航点值，使得无人车辆在每个航点可以根据下一个航点的航点值进行道路决策，实现了将复杂的道路决策优化问题进行简化，并且分两阶段计算航点值，第一阶段计算目标航点到目的地的全局代价，得到目标航点的航点值，第二阶段根据目标航点的航点值反向迭代计算目标航点到无人车辆的当前位置之间的航点的航点值，提高了航点值的计算速度，从而提高了道路决策效率，改善了现有技术采用模型预测控制的方法来获取最优道路决策，通过求解复杂的优化问题来得到最优道路决策，需要大量的运算能力来求解非线性优化问题，导致道路决策效率低的技术问题。
以上为本申请提供的一种道路决策方法的一个实施例，以下为本申请 提供的一种道路决策方法的另一个实施例。
本申请实施例提供的一种道路决策方法，包括：
步骤201、将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，可行驶道路至少包括一条车道，每条车道包括多个依次相连的航点。
步骤202、在无人车辆的当前位置到目的地之间的航点中确定目标航点，并计算目标航点到目的地的全局代价，得到目标航点的航点值。
步骤203、根据目标航点的航点值迭代计算目标航点到无人车辆的当前位置之间的航点的航点值。
步骤204、实时根据无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向目的地的当前道路决策结果。
步骤201至步骤204的具体内容与前述步骤101至步骤104的具体内容一致，在此不再进行赘述。
上述步骤是基于静态交通信息获取航点值以及进行道路决策，而无人车辆在实际行驶过程中所处的交通环境是动态的、随时间变化的，并且有多个其他交通参与者，这些交通参与者会动态地影响单步成本函数和无人车辆的转移模型，最终影响各航点的航点值。因此，在无人车辆行驶的过程中，需要根据交通信息更新航点值，进而更新道路决策结果。
进一步，本申请实施例中的道路决策方法还包括：
步骤205、根据交通信息更新道路决策结果。
具体更新过程为：
S2051、在无人车辆根据当前道路决策结果行驶时，获取受到交通信息影响的特殊航点。
在无人车辆根据当前道路决策结果行驶时，可以通过无人车辆上的传感器或车联网实时获取交通信息。
当交通信息包括静态交通参与者(停在路边的车辆、交通锥等)时，在无人车辆根据当前道路决策结果行驶时，根据静态交通参与者的位置确定受到静态交通参与者的影响的特殊航点。当静态交通参与者位于某个航点时，该航点即为特殊航点，请参考图5，在一个交通场景中，无人车辆在右侧车道上行驶，发现前方30米处有一交通锥封锁了右侧车道，无人车 辆可以预测到，无人车辆未来不能在航点S3通过保持车道行驶到航点S2，不能在航点S4通过右转变道行驶到航点S2，不能在航点S2通过保持车道行驶到航点S1，不能在航点S2通过左转变道至航点S0。即根据交通锥的位置，可以确定未来会受到交通锥影响的特殊航点为航点S2。若静态交通参与者位于两个航点之间时，例如，交通锥位于图5中航点S2与航点S1之间，无人车辆可以预测到，未来不能在航点S2通过保持车道行驶到航点S1，从而可以确定受影响的特殊航点为航点S2。
当交通信息包括动态交通参与者时，在无人车辆根据当前道路决策结果行驶时，从动态交通参与者中确定目标动态交通参与者；根据目标动态交通参与者的行驶速度和无人车辆的行驶速度确定受到目标动态交通参与者的影响的特殊航点。
无人车辆在行驶的过程中，会存在多个动态交通参与者(行人、行驶车辆等)的情况，若考虑所有的动态交通参与者的动态影响计算量非常大。为了减少计算量，提高航点值的更新速度，进而提高道路决策效率，本申请实施例中优选考虑无人车辆前方预置范围内、行驶速度低于车道限速的动态交通参与者。
进一步，从动态交通参与者中确定目标动态交通参与者的具体过程可以为：
将位于无人车辆前方预置范围内的动态交通参与者作为潜在目标动态交通参与者；
判断潜在目标动态交通参与者的行驶速度是否小于潜在目标动态交通参与者所在车道的限速值，得到判断结果；
根据潜在目标动态交通参与者的先验值和判断结果计算潜在目标动态交通参与者的置信度值；
基于置信度值从潜在目标动态交通参与者中确定目标交通参与者。
动态交通参与者的行为具有不确定性，在确定未来会受到目标动态交通参与者的影响的特殊航点时，需要确定考虑哪些目标交通参与者对航点值的动态影响。例如，如果无人车辆的前方车辆只缓慢行驶了1秒就开始加速，那么该前方车辆对航点值的影响较小，可以不考虑该前方车辆的影响，如果无人车辆的前方车辆缓慢行驶了一段时间，那么就需要考虑该前 方车辆对航点值的动态影响。
具体的，在确定潜在目标动态交通参与者后，可以给潜在目标动态交通参与者配置一个先验值，在得到潜在目标动态交通参与者的行驶速度是否小于潜在目标动态交通参与者所在车道的限速值的判断结果后，可以通过映射函数将判断结果映射为数值，例如，可以将潜在目标动态交通参与者的行驶速度小于所在车道的限速值的判断结果映射为数值1，将潜在目标动态交通参与者的行驶速度大于或等于所在车道的限速值的判断结果映射为数值0；然后通过预置权重系数对潜在目标动态交通参与者的先验值和判断结果对应的映射值进行加权求和，得到潜在目标动态交通参与者的置信度值。当潜在目标动态交通参与者在一段时间内的置信度值均大于预设置信度阈值时，则将该潜在目标动态交通参与者作为目标动态交通参与者，可以避免将突然加速或减速的潜在目标动态交通参与者作为目标动态交通参与者。
在确定目标动态交通参与者后，根据目标动态交通参与者的行驶速度和无人车辆的行驶速度确定未来会受到目标动态交通参与者的影响的特殊航点。请参考图7，在一个交通场景中，无人车辆(car1)以速度v 1匀速行驶，无人车辆的前方车辆(car2)以速度v 2匀速行驶，其中，v 1>v 2，假设根据静态交通信息计算的航点值得到无人车辆的当前决策结果为保持车道，即右侧车道是当前最好的车道。这些航点的航点值没有考虑缓慢行驶的前方车辆的动态影响，由于v 1>v 2，在未来某个区域(该区域由无人车辆和其前方车辆的速度差估计得到)，即图7中的阴影区域，无人车辆会靠近前方车辆，使得无人车辆在该阴影区域会受到缓慢行驶的前方车辆的影响，若无人车辆继续保持直行，则需要降低车速跟随前方车辆，即无人车辆由航点S2转移到航点S1时的短期代价会增加，进而影响航点S2的航点值，即航点S2为未来会受到前方车辆的影响的特殊航点。
S2052、当特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定特殊航点的下一个航点，并更新特殊航点到该下一个航点的短期代价。
当特殊航点为受静态交通参与者的影响的航点时，如图5所示，特殊航点为航点S2，根据当前道路决策结果(保持车道)可以确定航点S2的 下一个航点为航点S1，由于航点S2位置处有交通锥，使得无人车辆无法由航点S2到达航点S1，可以更新特殊航点S2到航点S1的短期代价C(S2,保持车道,S1)为一个较大的值(如50、100等)，具体的取值可以根据实际情况进行设置。
进一步，当特殊航点为受目标动态交通参与者的影响的航点时，特殊航点到对应的下一个航点的短期代价的更新过程为：
当特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定特殊航点的下一个航点，并确定特殊航点到该下一个航点的行驶距离；
根据行驶距离和目标交通参与者的行驶速度计算无人车辆从特殊航点到该下一个航点的短期代价，得到特殊航点到该下一个航点的更新后的短期代价。
以图7为例，根据当前道路决策结果(保持车道)可以确定特殊航点S2的下一个航点为航点S1，根据特殊航点S2和航点S1之间的行驶距离d以及无人车辆的前方车辆的行驶速度v 2可以计算特殊航点S2到航点S1的更新后的短期代价s/v 2，本申请实施例进一步考虑到目标动态交通参与者的动态影响会持续一定时间，因此，最终在特殊航点s到执行动作a转移到下一个航点s'的更新后的短期代价C(s,a,s')可以表示为：
C(s,a,s')＝β*(d s'-s/v)，
其中，β为截断参数，用于确定目标动态交通参与者的动态影响的持续时间，d s'-s为特殊航点s到对应的下一个航点s'的行驶距离，v为目标动态交通参与者的行驶速度。
S2053、基于特殊航点到该下一个航点更新后的短期代价更新特殊航点的航点值。
根据前述步骤可知，一个航点的航点值由该航点对应的下一个航点的航点值、航点之间转移的短期代价和状态转移概率计算得到，在更新了短期代价后，相应的航点值也会更新。可以理解的是，若状态转移概率更新了，相应的航点值也会进行更新。
以图5为例，假设在静态交通环境中，各航点之间转移的短期代价为1，变道成功率为20％，特殊航点S2转移到航点S1更新后的短期代价C(S2, 保持车道,S1)＝100，由于无法在特殊航点S2转移到航点S0，因此，特殊航点S2转移到航点S0的短期代价成本也会增加，假设特殊航点S2转移到航点S0更新后的短期代价成本C(S2,左转变道,S0)＝100。
若在特殊航点S2选择左转变道，更新后的航点值为V(S2) 左转变道＝(91+100)*20％+(51+100)*80％＝159；
若在特殊航点S2选择保持车道，更新后的航点值为V(S2) 保持车道＝(51+100)*100％＝151；
最终，特殊航点S2更新后的航点值为min(V(S2) 左转变道，V(S2) 保持车 道)＝151。
可以理解的是，若交通锥在航点S2和航点S1之间，即在特殊航点S2可以通过左转变道到航点S0，此时，特殊航点S2转移到航点S0的短期代价保持不变，即C(S2,左转变道,S0)＝1。此时，若在特殊航点S2选择左转变道，更新后的航点值为V(S2) 左转变道＝(91+1)*20％+(51+100)*80％＝139；最终，特殊航点S2更新后的航点值为min(V(S2) 左转变道，V(S2) 保持车道)＝139。
以图8为例，假设在静态交通环境中，各航点之间转移的短期代价为1，变道成功率为20％，根据静态交通信息计算得到的航点值如图8所示，假设计算得到的特殊航点S2转移到航点S1更新后的短期代价为30。
若在特殊航点S2选择左转变道，更新后的航点值为V(S2) 左转变道＝(84+1)*20％+(52+30)*80％＝83；
若在特殊航点S2选择保持车道，更新后的航点值为V(S2) 保持车道＝(52+30)*100％＝82；
最终，特殊航点S2更新后的航点值为min(V(S2) 左转变道，V(S2) 保持车道)＝82。
S2054、根据特殊航点更新后的航点值反向迭代更新特殊航点到无人车辆的当前位置之间的各航点的航点值，并返回步骤204。
根据图5可知，无人车辆无法在航点S3通过保持车道到达航点S2，因此，航点S3到特殊航点S2的短期代价也需要更新，假设更新后的短期代价C(S3,保持车道,S2)＝100。
若在航点S3选择左转变道，更新后的航点值为V(S3) 左转变道＝(84+1)*20％+(139+100)*80％＝208；
若在航点S3选择保持车道，更新后的航点值为V(S3) 保持车道 ＝(139+100)*100％＝239；
最终，航点S3更新后的航点值为min(V(S3) 左转变道，V(S3) 保持车道)＝208。
航点S4到特殊航点S2的短期代价也相应的需要更新，航点S4的航点值的更新过程与航点S3的航点值的更新过程类似，在此不再进行赘述。在航点S3和航点S4的航点值更新后，反向迭代更新航点S3、航点S4到无人车辆的当前位置之间的航点的航点值。需要说明的是，航点S3、航点S4到无人车辆的当前位置之间的航点对应的短期代价则保持不变。
对图5中的航点值进行更新后，得到的更新后的航点值如图6所示，根据更新后的航点值可知，无人车辆在当前航点将左转变道至左侧车道，进而超越交通锥。对图8中的航点值进行更新后，得到的更新后的航点值如图9所示，根据图9更新后的航点值可知，无人车辆将左转变道至左侧车道超越前方慢车。
在静态交通环境中各航点的航点值的计算不考虑时间的，即不考虑动态的交通环境的影响。当无人车辆前方有一行驶极慢的动态交通参与者时，此时该无人车辆需要付出巨大的时间从当前航点前进至前方的下一个航点，即无人车辆在各航点之间转移所付出的短期代价与交通环境紧密相关，短期代价会根据每一帧的交通信息更新，是动态变化的，相应的，航点值也是动态变化的。本申请实施例中，特殊航点到无人车辆的当前位置之间的各航点的航点值的更新公式可以表示为：
Figure PCTCN2022119830-appb-000001
式中，V(s)为航点s更新后的航点值，C t(s,a,s′)为无人车辆在当前时刻t从航点s执行动作a到达航点s′的短期代价，V(s′)为航点s′的航点值，A为无人车辆在航点s的可执行动作集合，
Figure PCTCN2022119830-appb-000002
为基于时变转移模型T t和在当前时刻t的交通参与者集合
Figure PCTCN2022119830-appb-000003
的期望值函数。
由于有其他交通参与者的存在，转移模型变得与时间有关。在各时刻，无人车辆的当前航点(即当前状态)是已知的，无人车辆选择某一可执行的动作(左转变道、右转变道或保持车道)所到达的状态是不确定的，例如，目标变道车道的交通密度接近其容量，或者该目标变道车道的后方车辆正在迅速接近，无人车辆即使做出了变道的动作，也不一定能成功变道到该目标变道车道。因此，需要通过观测无人车辆周围的交通信息，动态 更新各航点之间转移的变道成功率。其中，可执行的动作由无人驾驶车辆所在的车道决定，例如，无人车辆在最右车道，该无人车辆右方没有可行驶道路，此时右转变道是不可执行的动作，直行和左转变道为可执行的动作。
在本申请实施例中，对于无人车辆预置范围之外的航点，无人车辆预置范围之外的航点之间的变道成功率P(succ. t＝1)继承在静态交通环境中计算得到的变道成功率P 0，即P(succ. t＝1)＝P 0；对于无人车辆预置范围之内的航点，则根据交通信息更新无人车辆预置范围内的航点之间转移时的变道成功率。
具体的，根据交通信息获取无人车辆的后侧方车辆与无人车辆的当前距离和无人车辆的后侧方车辆的当前让步概率，更新无人车辆在当前航点的变道成功率。
对于无人车辆在当前航点的变道成功率，需要考虑无人车辆的后侧方车辆与无人车辆的当前距离d t，以及无人车辆的后侧方车辆的当前让步概率P(succ. t＝1|y t)(受后侧方车辆的让步意愿y t的影响)，即无人车辆在当前航点的变道成功率P(succ. t＝1|d t,y t)可以表示为：
P(succ. t＝1|d t,y t)∝P(succ. t＝1|d t)·P(succ. t＝1|y t)；
其中，P(succ. t＝1|d t)用于根据无人车辆与后侧方车辆的当前距离控制换道成功率，P(succ. t＝1|y t)用于根据后侧方车辆的配合情况来控制变道成功率，∝为正比符号。
进一步，P(succ. t＝1|d t)的计算公式可以为：
Figure PCTCN2022119830-appb-000004
式中，P 0为在静态交通环境中计算得到的在当前航点的变道成功率，即当前航点更新前的变道成功率；d safe为安全变道距离，当d t＝d safe时，P(succ. t＝1|d t)＝P 0。
进一步，无人车辆的后侧方车辆的当前让步概率的计算过程为：
根据无人车辆的后侧方车辆的当前加速度和该后侧方车辆在前一时刻的让步概率计算该后侧方车辆的当前让步概率，其中，该后侧方车辆的初始让步概率通过初始化得到。P(succ. t＝1|y t)的计算公式可以表示为：
P(succ. t＝1|y t)＝αP(succ. t-1＝1|y t-1)+(1-α)ΙΙ(a t＜0)；
式中，P(succ. t＝1|y t)为无人车辆的后侧方车辆的当前让步概率，P(succ. t-1＝1|y t-1)为后侧方车辆在前一时刻的让步概率，α为更新率，a t为后侧方车辆的当前加速度，ΙΙ(*)为映射函数，当事件*为真时，ΙΙ(*)＝1，当事件*为假时，ΙΙ(*)＝0，即当a t<0时，ΙΙ(a t＜0)＝1，当a t≥0时，ΙΙ(a t＜0)＝0。
后侧方车辆的初始让步概率通过初始化得到，不同的后侧方车辆的初始让步概率可以为相同的初始值，在行驶过程中可以根据后侧方车辆的反应更新该后侧方车辆的让步概率。
对于无人车辆预置范围内的剩余航点，即无人车辆预置范围内的除无人车辆所在的当前航点之外的其他航点，根据目标变道车道的交通密度ρ t更新在无人车辆预置范围内的剩余航点的变道成功率，目标变道车道为变道后的车道，可以表示为：
Figure PCTCN2022119830-appb-000005
式中，P(succ. t＝1|ρ t)为无人车辆预置范围内的剩余航点在时刻t的交通密度下的变道成功率，β为衰减因子，ρ t为目标变道车道在时刻t的交通密度，δ为目标变道车道的通行能力，P max为变道成功率阈值。
进一步，当受到静态交通参与者影响的特殊航点为根据当前道路决策结果行驶未来会到达的航点，且特殊航点所在车道的相邻车道无法通行时，本申请实施例中的方法还包括：
将受到静态交通参与者影响的特殊航点所在车道与其相邻车道之间的车道分隔线划分为若干个依次相连的航点；根据车道分割线的相邻车道上的航点的航点值、各航点之间转移的短期代价和状态转移概率计算该车道分隔线上各航点的航点值，并返回步骤204。其中，车道分割线上的航点之间转移的短期代价会比在正常车道上的航点之间转移的短期代价要高，具体取值可以根据实际情况进行设置。
例如，如图10所示，在一种交通场景中，无人车辆前方有两个车道，无人车辆在右车道上行驶，右车道前方有一个交通锥，而左车道通向死胡同，使得计算得到的左车道上各航点的航点值比右车道上的航点的航点值 高很多，即无人车辆从右车道变道到左车道的短期代价很高，而右车道前方又有交通锥，无法一直保持直行，在这种情况下，可以将左车道和右车道之间的车道分隔线(即图10中的实线)划分为若干个依次相连的航点，然后通过上述步骤S2053中的航点值更新公式计算该车道分隔线上的航点的航点值，当分割线上的航点的航点值小于受交通锥影响的右车道更新后的航点值和左车道的航点的航点值时，无人车辆可以以较少的代价变道到车道分割上的航点来超越交通锥。其中，在静态交通环境下，假设车道上各航点之间转移的短期代价为1，车道分割线上的航点之间转移的短期代价为30，由于交通锥的影响，右侧车道上的特殊航点到下一个航点的更新后的短期代价为100，基于此计算得到的车道分割上的航点的航点值如图10所示，在该场景下，无人车辆保持直行一段时间后，将变道至车道分隔线上行驶，以超越交通锥。
本申请实施例考虑到，若采用模型预测控制的方法来获取最优道路决策，需要通过求解复杂的优化问题来得到最优道路决策，需要大量的运算能力来求解非线性优化问题，严重依赖于环境模型的构建，难以被有效应用到无人驾驶车辆的决策***中。而本申请实施例分两部分来求解优化问题，一部分是通过对目标航点的全局搜索来获取目标航点的全局代价，另一部分是通过观测实时交通信息，动态修正在不同状态之间转移所付出的短期代价与变道成功率，将高维度多智能体的优化问题简化为低维度单一智能体的优化问题，求解速度更快。通过对无人驾驶车辆的可行道路的全局代价与短期代价进行快速实时量化分析，对道路的短期代价与全局代价进行平衡，使得无人驾驶车辆可以以少量的运算获取最优道路决策结果，从而在最优的时间进行服从全局导航主动换道、主动换道超慢车、主动换道脱离潜在风险区域(如施工区域、交通事故区域等)、主动换道躲避优先车辆(如警车、救护车等)等。
以上为本申请提供的一种道路决策方法的另一个实施例，以下为本申请提供的一种道路决策***的一个实施例。
请参考图11，本申请实施例提供的一种道路决策***，包括：
划分模块，用于将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，可行驶道路至少包括一条车道，每条车道包括多个依次 相连的航点；
第一计算模块，用于在无人车辆的当前位置到目的地之间的航点中确定目标航点，并计算目标航点到目的地的全局代价，得到目标航点的航点值；
第二计算模块，用于根据目标航点的航点值迭代计算目标航点到无人车辆的当前位置之间的航点的航点值；
决策模块，用于实时根据无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向目的地的当前道路决策结果，动作为左转变道、保持车道或右转变道。
作为进一步地改进，第一计算模块具体用于：
通过图搜索算法获取目标航点到目的地的最短路径；
基于最短路径和预置行驶速度计算无人车辆从目标航点到目的地的行驶时间；
基于无人车辆从目标航点到目的地的行驶时间获取目标航点到目的地的全局代价；
将目标航点到目的地的全局代价作为目标航点的航点值。
作为进一步地改进，本申请实施例中的道路决策***，还包括：航点值更新模块，用于：
在无人车辆根据当前道路决策结果行驶时，获取受到交通信息影响的特殊航点；
当特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定特殊航点的下一个航点，并更新特殊航点到该下一个航点的短期代价；
基于特殊航点到该下一个航点更新后的短期代价更新特殊航点的航点值；
根据特殊航点更新后的航点值反向迭代更新特殊航点到无人车辆的当前位置之间的各航点的航点值，并触发决策模块。
作为进一步地改进，本申请实施例中的道路决策***，还包括：第三计算模块，用于：
当受到静态交通参与者影响的特殊航点为根据当前道路决策结果行驶 未来会到达的航点，且特殊航点所在车道的相邻车道无法通行时，将受到静态交通参与者影响的特殊航点所在车道与该相邻车道之间的车道分隔线划分为若干个依次相连的航点；
根据车道分割线的相邻车道上的航点的航点值、各航点之间转移的短期代价和状态转移概率计算该车道分隔线上各航点的航点值，并触发决策模块。
作为进一步地改进，状态转移概率包括变道成功率，本申请实施例中的道路决策***还包括：
变道成功率更新模块，用于根据交通信息更新无人车辆预置范围内的航点之间转移时的变道成功率。
作为进一步地改进，变道成功率更新模块具体用于：
根据交通信息获取无人车辆的后侧方车辆与无人车辆的当前距离和无人车辆的后侧方车辆的当前让步概率，更新无人车辆在当前航点的变道成功率；
根据目标变道车道的交通密度更新在无人车辆预置范围内的剩余航点的变道成功率，目标变道车道为变道后的车道，无人车辆预置范围内的剩余航点为无人车辆预置范围内的除无人车辆所在的当前航点之外的其他航点。
本申请实施例中，将无人车辆到目的地之间的可行驶道路划分为航点，通过计算各航点的航点值，使得无人车辆在每个航点可以根据下一个航点的航点值进行道路决策，实现了将复杂的道路决策优化问题进行简化，并且分两阶段计算航点值，第一阶段计算目标航点到目的地的全局代价，得到目标航点的航点值，第二阶段根据目标航点的航点值反向迭代计算目标航点到无人车辆的当前位置之间的航点的航点值，提高了航点值的计算速度，从而提高了道路决策效率，改善了现有技术采用模型预测控制的方法来获取最优道路决策，通过求解复杂的优化问题来得到最优道路决策，需要大量的运算能力来求解非线性优化问题，导致道路决策效率低的技术问题。
本申请实施例还提供了一种道路决策设备，设备包括处理器以及存储器；
存储器用于存储程序代码，并将程序代码传输给处理器；
处理器用于根据程序代码中的指令执行前述方法实施例中的道路决策方法。
本申请实施例还提供了一种计算机可读存储介质，计算机可读存储介质用于存储程序代码，程序代码被处理器执行时实现前述方法实施例中的道路决策方法。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的***，装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
本申请的说明书及上述附图中的术语“第一”、“第二”、“第三”、“第四”等(如果存在)是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例例如能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、***、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
应当理解，在本申请中，“至少一个(项)”是指一个或者多个，“多个”是指两个或两个以上。“和/或”，用于描述关联对象的关联关系，表示可以存在三种关系，例如，“A和/或B”可以表示：只存在A，只存在B以及同时存在A和B三种情况，其中A，B可以是单数或者复数。字符“/”一般表示前后关联对象是一种“或”的关系。“以下至少一项(个)”或其类似表达，是指这些项中的任意组合，包括单项(个)或复数项(个)的任意组合。例如，a，b或c中的至少一项(个)，可以表示：a，b，c，“a和b”，“a和c”，“b和c”，或“a和b和c”，其中a，b，c可以是单个，也可以是多个。
在本申请所提供的几个实施例中，应该理解到，所揭露的装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以 有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个***，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以通过一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(英文全称：Read-Only Memory，英文缩写：ROM)、随机存取存储器(英文全称：Random Access Memory，英文缩写：RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，以上实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围。

Claims (17)

  1. 一种道路决策方法，其特征在于，包括：
    将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，所述可行驶道路至少包括一条车道，每条车道包括多个依次相连的航点；
    在所述无人车辆的当前位置到所述目的地之间的航点中确定目标航点，并计算所述目标航点到所述目的地的全局代价，得到所述目标航点的航点值；
    根据所述目标航点的航点值迭代计算所述目标航点到所述无人车辆的当前位置之间的航点的航点值；
    实时根据所述无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向所述目的地的当前道路决策结果。
  2. 根据权利要求1所述的道路决策方法，其特征在于，所述计算所述目标航点到所述目的地的全局代价，得到所述目标航点的航点值，包括：
    通过图搜索算法获取所述目标航点到所述目的地的最短路径；
    基于所述最短路径和预置行驶速度计算所述无人车辆从所述目标航点到所述目的地的行驶时间；
    基于所述无人车辆从所述目标航点到所述目的地的行驶时间获取所述目标航点到所述目的地的全局代价；
    将所述目标航点到所述目的地的全局代价作为所述目标航点的航点值。
  3. 根据权利要求2所述的道路决策方法，其特征在于，所述基于所述无人车辆从所述目标航点到所述目的地的行驶时间获取所述目标航点到所述目的地的全局代价，包括：
    将所述无人车辆从所述目标航点到所述目的地的行驶时间作为所述目标航点到所述目的地的全局代价；
    或，根据所述目标航点到所述目的地的目标信息和所述无人车辆从所述目标航点到所述目的地的行驶时间计算所述目标航点到所述目的地的全局代价。
  4. 根据权利要求1所述的道路决策方法，其特征在于，在根据所述目标航点的航点值迭代计算所述目标航点到所述无人车辆的当前位置之间的航点的航点值时，所述目标航点对应的上一个航点的航点值的计算过程为：
    计算所述目标航点对应的上一个航点转移到所述目标航点的短期代价和状态转移概率；
    在所述目标航点的航点值的基础上叠加所述目标航点对应的上一个航点转移到所述目标航点的所述短期代价，并结合所述状态转移概率计算所述目标航点对应的上一个航点的航点值。
  5. 根据权利要求1所述的道路决策方法，其特征在于，所述无人车辆的当前位置到所述目的地之间设置有多个所述目标航点。
  6. 根据权利要求4所述的道路决策方法，其特征在于，所述方法还包括：
    在所述无人车辆根据当前道路决策结果行驶时，获取受到交通信息影响的特殊航点；
    当所述特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定所述特殊航点的下一个航点，并更新所述特殊航点到该下一个航点的短期代价；
    基于所述特殊航点到该下一个航点更新后的短期代价更新所述特殊航点的航点值；
    根据所述特殊航点更新后的航点值反向迭代更新所述特殊航点到所述无人车辆的当前位置之间的各航点的航点值，并返回所述实时根据所述无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向所述目的地的当前道路决策结果的步骤。
  7. 根据权利要求6所述的道路决策方法，其特征在于，当所述交通信息包括静态交通参与者时；
    所述在所述无人车辆根据当前道路决策结果行驶时，获取受到交通信息影响的特殊航点，包括：
    在所述无人车辆根据当前道路决策结果行驶时，根据所述静态交通参与者的位置确定受到所述静态交通参与者的影响的特殊航点。
  8. 根据权利要求7所述的道路决策方法，其特征在于，所述方法还包 括：
    当受到所述静态交通参与者影响的所述特殊航点为根据当前道路决策结果行驶未来会到达的航点，且所述特殊航点所在车道的相邻车道无法通行时，将受到所述静态交通参与者影响的所述特殊航点所在车道与该相邻车道之间的车道分隔线划分为若干个依次相连的航点；
    根据所述车道分割线的相邻车道上的航点的航点值、各航点之间转移的短期代价和状态转移概率计算该车道分隔线上各航点的航点值，并返回所述实时根据所述无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向所述目的地的当前道路决策结果的步骤。
  9. 根据权利要求6所述的道路决策方法，其特征在于，当所述交通信息包括动态交通参与者时；
    所述在所述无人车辆根据当前道路决策结果行驶时，获取受到交通信息影响的特殊航点，包括：
    在所述无人车辆根据当前道路决策结果行驶时，从所述动态交通参与者中确定目标动态交通参与者；
    根据所述目标动态交通参与者的行驶速度和所述无人车辆的行驶速度确定受到所述目标动态交通参与者的影响的特殊航点。
  10. 根据权利要求9所述的道路决策方法，其特征在于，所述从所述动态交通参与者中确定目标动态交通参与者，包括：
    将位于所述无人车辆前方预置范围内的动态交通参与者作为潜在目标动态交通参与者；
    判断所述潜在目标动态交通参与者的行驶速度是否小于所述潜在目标动态交通参与者所在车道的限速值，得到判断结果；
    根据所述潜在目标动态交通参与者的值和所述判断结果计算所述潜在目标动态交通参与者的置信度值；
    基于所述置信度值从所述潜在目标动态交通参与者中确定目标交通参与者。
  11. 根据权利要求9所述的道路决策方法，其特征在于，所述当所述特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定所述特殊航点的下一个航点，并更新所述特殊航点到该下 一个航点的短期代价，包括：
    当所述特殊航点为根据当前道路决策结果行驶未来会到达的航点时，根据当前道路决策结果确定所述特殊航点的下一个航点，并确定所述特殊航点到该下一个航点的行驶距离；
    根据所述行驶距离和所述目标交通参与者的行驶速度计算所述无人车辆从所述特殊航点到该下一个航点的短期代价，得到所述特殊航点到该下一个航点的更新后的短期代价。
  12. 根据权利要求6所述的道路决策方法，其特征在于，所述状态转移概率包括变道成功率，所述方法还包括：
    根据所述交通信息更新所述无人车辆预置范围内的航点之间转移时的变道成功率。
  13. 根据权利要求12所述的道路决策方法，其特征在于，所述根据所述交通信息更新所述无人车辆预置范围内的航点之间转移时的变道成功率，包括：
    根据所述交通信息获取所述无人车辆的后侧方车辆与所述无人车辆的当前距离和所述无人车辆的后侧方车辆的当前让步概率，更新所述无人车辆在当前航点的变道成功率；
    根据目标变道车道的交通密度更新在所述无人车辆预置范围内的剩余航点的变道成功率，所述目标变道车道为变道后的车道，所述无人车辆预置范围内的剩余航点为所述无人车辆预置范围内的除所述无人车辆所在的当前航点之外的其他航点。
  14. 根据权利要求13所述的道路决策方法，其特征在于，所述无人车辆的后侧方车辆的当前让步概率的计算过程为：
    根据所述无人车辆的后侧方车辆的当前加速度和该后侧方车辆在前一时刻的让步概率计算该后侧方车辆的当前让步概率，其中，该后侧方车辆的初始让步概率通过初始化得到。
  15. 一种道路决策***，其特征在于，包括：
    划分模块，用于将无人车辆的当前位置到目的地之间的可行驶道路划分为若干个航点，所述可行驶道路至少包括一条车道，每条车道包括多个依次相连的航点；
    第一计算模块，用于在所述无人车辆的当前位置到所述目的地之间的航点中确定目标航点，并计算所述目标航点到所述目的地的全局代价，得到所述目标航点的航点值；
    第二计算模块，用于根据所述目标航点的航点值迭代计算所述目标航点到所述无人车辆的当前位置之间的航点的航点值；
    决策模块，用于实时根据所述无人车辆在当前位置对应的下一个航点的航点值确定在当前位置的动作，得到驶向所述目的地的当前道路决策结果。
  16. 一种道路决策设备，其特征在于，所述设备包括处理器以及存储器；
    所述存储器用于存储程序代码，并将所述程序代码传输给所述处理器；
    所述处理器用于根据所述程序代码中的指令执行权利要求1-14任一项所述的道路决策方法。
  17. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质用于存储程序代码，所述程序代码被处理器执行时实现权利要求1-14任一项所述的道路决策方法。
PCT/CN2022/119830 2021-09-29 2022-09-20 一种道路决策方法、***、设备和介质 WO2023051312A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202111155395.9 2021-09-29
CN202111155395.9A CN113911136B (zh) 2021-09-29 2021-09-29 一种无人驾驶车辆变道决策方法、***、设备和介质
CN202111453193.2A CN113928341B (zh) 2021-11-30 2021-11-30 一种道路决策方法、***、设备和介质
CN202111453193.2 2021-11-30

Publications (1)

Publication Number Publication Date
WO2023051312A1 true WO2023051312A1 (zh) 2023-04-06

Family

ID=85781292

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/119830 WO2023051312A1 (zh) 2021-09-29 2022-09-20 一种道路决策方法、***、设备和介质

Country Status (1)

Country Link
WO (1) WO2023051312A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116853288A (zh) * 2023-06-21 2023-10-10 广州汽车集团股份有限公司 车辆驾驶方法、装置、车辆及存储介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100332116A1 (en) * 2009-06-25 2010-12-30 Denso International America, Inc. Off road navigation system
CN106940933A (zh) * 2017-03-08 2017-07-11 北京理工大学 一种基于智能交通***的智能车辆决策换道方法
CN109976334A (zh) * 2019-02-25 2019-07-05 广州文远知行科技有限公司 车辆变道方法、装置、设备和存储介质
CN112747744A (zh) * 2020-12-22 2021-05-04 浙江大学 一种结合航位推算和多车道路网地图的车辆定位方法
CN113911136A (zh) * 2021-09-29 2022-01-11 广州文远知行科技有限公司 一种无人驾驶车辆变道决策方法、***、设备和介质
CN113928341A (zh) * 2021-11-30 2022-01-14 广州文远知行科技有限公司 一种道路决策方法、***、设备和介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100332116A1 (en) * 2009-06-25 2010-12-30 Denso International America, Inc. Off road navigation system
CN106940933A (zh) * 2017-03-08 2017-07-11 北京理工大学 一种基于智能交通***的智能车辆决策换道方法
CN109976334A (zh) * 2019-02-25 2019-07-05 广州文远知行科技有限公司 车辆变道方法、装置、设备和存储介质
CN112747744A (zh) * 2020-12-22 2021-05-04 浙江大学 一种结合航位推算和多车道路网地图的车辆定位方法
CN113911136A (zh) * 2021-09-29 2022-01-11 广州文远知行科技有限公司 一种无人驾驶车辆变道决策方法、***、设备和介质
CN113928341A (zh) * 2021-11-30 2022-01-14 广州文远知行科技有限公司 一种道路决策方法、***、设备和介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116853288A (zh) * 2023-06-21 2023-10-10 广州汽车集团股份有限公司 车辆驾驶方法、装置、车辆及存储介质
CN116853288B (zh) * 2023-06-21 2024-04-02 广州汽车集团股份有限公司 车辆驾驶方法、装置、车辆及存储介质

Similar Documents

Publication Publication Date Title
US20220126822A1 (en) Vehicle driving control method and apparatus
CN111656295B (zh) 用于自主运载工具的对象交互预测***及方法
WO2021073389A1 (zh) 车辆自动控制方法及变道意图预测网络的训练方法
CN109791409B (zh) 自主车辆的运动控制决策
US20210197720A1 (en) Systems and methods for incident detection using inference models
US20210197813A1 (en) Systems and methods for appropriate speed inference
JP7060625B2 (ja) 自動運転車において３ｄｃｎｎネットワークを用いてソリューション推断を行うｌｉｄａｒ測位
WO2020034903A1 (zh) 一种基于拓扑图的智能导航的方法和***
CN109489675A (zh) 用于自动驾驶车辆的基于成本的路径规划
CN109491377A (zh) 用于自动驾驶车辆的基于dp和qp的决策和规划
US11354913B1 (en) Systems and methods for improving vehicle predictions using point representations of scene
CN109491376A (zh) 用于自动驾驶车辆的基于动态规划和梯度下降的决策和规划
CN108205830A (zh) 识别对于无人驾驶车辆的个人驾驶偏好的方法和***
JP2023510136A (ja) 知覚、予測又は計画のための地理的位置特定モデル
CN109491375A (zh) 用于自动驾驶车辆的基于驾驶场景的路径规划
CN109643118B (zh) 基于关于车辆的环境的与功能相关的信息来影响车辆的功能
CN108139756A (zh) 为自动驾驶车辆构建周围环境以制定驾驶决策的方法和***
CN113682318B (zh) 车辆行驶控制方法及装置
US20210402991A1 (en) Object Trajectory Association and Tracking
WO2022222401A1 (zh) 自主泊车的方法、装置、设备及自动驾驶车辆
WO2023051312A1 (zh) 一种道路决策方法、***、设备和介质
CN115344052B (zh) 基于改进的群优化算法的车辆路径控制方法及控制***
CN111566446A (zh) 运输机路线选择***
CN115083175A (zh) 基于车路协同的信号管控方法、相关装置及程序产品
CN117406756B (zh) 一种运动轨迹参数的确定方法、装置、设备和存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22874699

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18560390

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE