CN105059213B - A kind of intelligence is with vehicle control and method - Google Patents
A kind of intelligence is with vehicle control and method Download PDFInfo
- Publication number
- CN105059213B CN105059213B CN201510489170.5A CN201510489170A CN105059213B CN 105059213 B CN105059213 B CN 105059213B CN 201510489170 A CN201510489170 A CN 201510489170A CN 105059213 B CN105059213 B CN 105059213B
- Authority
- CN
- China
- Prior art keywords
- neutral net
- actual spacing
- vehicle
- neural network
- car
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 230000002708 enhancing effect Effects 0.000 claims abstract description 87
- 238000001514 detection method Methods 0.000 claims abstract description 15
- 230000008859 change Effects 0.000 claims abstract description 11
- 230000007935 neutral effect Effects 0.000 claims description 105
- 230000009471 action Effects 0.000 claims description 55
- 238000013528 artificial neural network Methods 0.000 claims description 46
- 238000011156 evaluation Methods 0.000 claims description 29
- 238000011478 gradient descent method Methods 0.000 claims description 14
- WHGYBXFWUBPSRW-FOUAGVGXSA-N beta-cyclodextrin Chemical compound OC[C@H]([C@H]([C@@H]([C@H]1O)O)O[C@H]2O[C@@H]([C@@H](O[C@H]3O[C@H](CO)[C@H]([C@@H]([C@H]3O)O)O[C@H]3O[C@H](CO)[C@H]([C@@H]([C@H]3O)O)O[C@H]3O[C@H](CO)[C@H]([C@@H]([C@H]3O)O)O[C@H]3O[C@H](CO)[C@H]([C@@H]([C@H]3O)O)O3)[C@H](O)[C@H]2O)CO)O[C@@H]1O[C@H]1[C@H](O)[C@@H](O)[C@@H]3O[C@@H]1CO WHGYBXFWUBPSRW-FOUAGVGXSA-N 0.000 claims description 10
- 210000004218 nerve net Anatomy 0.000 claims description 2
- 230000001133 acceleration Effects 0.000 abstract description 6
- 230000006870 function Effects 0.000 description 47
- 230000013016 learning Effects 0.000 description 13
- 238000005728 strengthening Methods 0.000 description 8
- 230000003044 adaptive effect Effects 0.000 description 5
- 230000007613 environmental effect Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R16/00—Electric or fluid circuits specially adapted for vehicles and not otherwise provided for; Arrangement of elements of electric or fluid circuits specially adapted for vehicles and not otherwise provided for
- B60R16/02—Electric or fluid circuits specially adapted for vehicles and not otherwise provided for; Arrangement of elements of electric or fluid circuits specially adapted for vehicles and not otherwise provided for electric constitutive elements
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60K—ARRANGEMENT OR MOUNTING OF PROPULSION UNITS OR OF TRANSMISSIONS IN VEHICLES; ARRANGEMENT OR MOUNTING OF PLURAL DIVERSE PRIME-MOVERS IN VEHICLES; AUXILIARY DRIVES FOR VEHICLES; INSTRUMENTATION OR DASHBOARDS FOR VEHICLES; ARRANGEMENTS IN CONNECTION WITH COOLING, AIR INTAKE, GAS EXHAUST OR FUEL SUPPLY OF PROPULSION UNITS IN VEHICLES
- B60K31/00—Vehicle fittings, acting on a single sub-unit only, for automatically controlling vehicle speed, i.e. preventing speed from exceeding an arbitrarily established velocity or maintaining speed at a particular velocity, as selected by the vehicle operator
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60K—ARRANGEMENT OR MOUNTING OF PROPULSION UNITS OR OF TRANSMISSIONS IN VEHICLES; ARRANGEMENT OR MOUNTING OF PLURAL DIVERSE PRIME-MOVERS IN VEHICLES; AUXILIARY DRIVES FOR VEHICLES; INSTRUMENTATION OR DASHBOARDS FOR VEHICLES; ARRANGEMENTS IN CONNECTION WITH COOLING, AIR INTAKE, GAS EXHAUST OR FUEL SUPPLY OF PROPULSION UNITS IN VEHICLES
- B60K2310/00—Arrangements, adaptations or methods for cruise controls
- B60K2310/24—Speed setting methods
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Transportation (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a kind of intelligence with vehicle control and method, belong to automobile active safety technical field.The intelligence includes with vehicle control:Detection module, the actual spacing for detecting vehicle and front vehicles in real time, and enhancing signal corresponding with the actual spacing is produced according to the actual spacing, the enhancing signal is for representing deviation amplitude of the dream car of the actual spacing and setting away between;Strengthen study module, for by the way of enhancing study, according to the actual spacing and the enhancing signal, determining the adjustment amplitude of the travel speed of the vehicle;With car module, for according to the adjustment amplitude, the travel speed of the vehicle being adjusted, to change the actual spacing.The situation for constantly switching between acceleration and deceleration and causing vehicle to jolt is not present in the present invention, and stability, reliability and comfortableness with car are improved.
Description
Technical field
The present invention relates to automobile active safety technical field, more particularly to a kind of intelligence is with vehicle control and method.
Background technology
Urban traffic blocking gives people trip and brings inconvenience, while bringing serious social concern and environmental problem.
For example, during traffic congestion, driver needs constantly to switch the state of automobile between transport condition and dead ship condition, causes driver
Spiritual high concentration, driving fatigue, improve the possibility of traffic accident generation.
Intelligence frees in can making driving fatigue of the driver during the traffic congestion with the application of driving skills art, it is to avoid traffic
The generation of accident.Intelligence is mainly the reality using the devices such as camera, radar detection vehicle and front vehicles with driving skills art at present
Border spacing, and by the way of supervised learning, the traveling speed of vehicle is adjusted according to the difference of actual spacing and the following distance of setting
Degree, minimizes the deviation between actual spacing and the following distance of setting.
During the present invention is realized, inventor has found that prior art at least has problems with:
When actual spacing be more than setting following distance when, if front vehicles due to emergency situations (as turned) just
Reduced Speed Now, then accelerate the travel speed of vehicle according to the difference of actual spacing and the following distance of setting, can cause actual spacing
It is less than the following distance of setting again.Now according still further to the traveling speed of actual spacing and the subtractive slow train of the following distance of setting
Degree, if front vehicles (as completed to turn) are given it the gun again, can cause actual spacing but also more than the following distance of setting not only.Such as
This repeatedly, vehicle constantly accelerates and slowed down, and causes vehicle to jolt, and the stability, reliability and comfortableness with car are poor.
The content of the invention
In order to solve prior art with car stability, reliability and poor comfortableness the problem of, the embodiment of the present invention is carried
A kind of intelligence has been supplied with vehicle control and method.The technical scheme is as follows:
On the one hand, the embodiments of the invention provide a kind of intelligence with vehicle control, the intelligence is with vehicle control bag
Include:
Detection module, is produced for detecting vehicle and the actual spacing of front vehicles in real time, and according to the actual spacing
Enhancing signal corresponding with the actual spacing, the enhancing signal be used to representing the dream car of the actual spacing and setting away from
Between deviation amplitude;
Strengthen study module, by the way of learning using enhancing, according to the actual spacing and the enhancing signal, really
The adjustment amplitude of the travel speed of the fixed vehicle;
With car module, for according to the adjustment amplitude, the travel speed of the vehicle being adjusted, to change the actual car
Away from;
The enhancing study module, including:
Neutral net is acted, for according to the actual spacing, producing the adjustment amplitude of the travel speed of the vehicle;
Neutral net is evaluated, for according to the actual spacing, the enhancing signal and the adjustment amplitude, producing
Cost function, the cost function strengthens the approximate representation of signal to be described;According to the cost function, the evaluation god is adjusted
Neural network weight through network, to minimize the error of the cost function and the enhancing signal;
The action neutral net is additionally operable to, what the evaluation neutral net after being adjusted according to neural network weight was produced
The cost function, adjusts the neural network weight of the action neutral net, optimal adjustment amplitude is obtained, to minimize
State the error of cost function and desired value, the desired value be the actual spacing reach the dream car away from when produce described in
Cost function.
Alternatively, the action neutral net and the evaluation neutral net use Nonlinear Multi perceptron.
Alternatively, the action neutral net is used for,
Then set the regulation of number of times to the neural network weight of the action neutral net using gradient descent method;
The evaluation neutral net is used for,
Then set the regulation of number of times to the neural network weight of the evaluation neutral net using gradient descent method.
In a kind of possible implementation of the present invention, the adjustment amplitude symbolization function.
On the other hand, the embodiments of the invention provide a kind of intelligence with car control method, the intelligence is with car control method
Including:
Detection vehicle and the actual spacing of front vehicles, and produced and the actual spacing according to the actual spacing in real time
Corresponding enhancing signal, the enhancing signal is for representing deviation width of the dream car of the actual spacing and setting away between
Degree;
By the way of enhancing study, according to the actual spacing and the enhancing signal, the traveling of the vehicle is determined
The adjustment amplitude of speed;
According to the adjustment amplitude, the travel speed of the vehicle is adjusted, to change the actual spacing;
It is described that the adjustment amplitude of the travel speed of the vehicle is determined according to the actual spacing and the enhancing signal,
Including:
Using action neutral net according to the actual spacing, the adjustment amplitude of the travel speed of the vehicle is produced;
Using neutral net is evaluated according to the actual spacing, the enhancing signal and the adjustment amplitude, generation is produced
Valency function, the cost function strengthens the approximate representation of signal to be described;
According to the cost function, the neural network weight for evaluating neutral net is adjusted, to minimize the cost
The error of function and the enhancing signal;
The cost function that the evaluation neutral net after being adjusted according to neural network weight is produced, adjusts described dynamic
Make the neural network weight of neutral net, obtain optimal adjustment amplitude, to minimize the mistake of the cost function and desired value
Difference, the desired value be the actual spacing reach the dream car away from when the cost function that produces.
Alternatively, the action neutral net and the evaluation neutral net use Nonlinear Multi perceptron.
Alternatively, it is described according to the cost function, the neural network weight for evaluating neutral net is adjusted, including:
Then set the regulation of number of times to the neural network weight of the action neutral net using gradient descent method;
It is described adjusted according to neural network weight after the evaluation neutral net produce the cost function, adjust institute
The neural network weight of action neutral net is stated, including:
Then set the regulation of number of times to the neural network weight of the evaluation neutral net using gradient descent method.
In a kind of possible implementation of the present invention, the adjustment amplitude symbolization function.
The beneficial effect that technical scheme provided in an embodiment of the present invention is brought is:
By using the mode of enhancing study, according to actual spacing and enhancing signal, the tune of the travel speed of vehicle is determined
View picture degree, enhancing signal, can be in travel speed for representing deviation amplitude of the dream car of actual spacing and setting away between
Adjustment during, using adaptive dynamic programming method, constantly by strengthening the deviation amplitude that signal reflects, autonomous learning is such as
What determines suitable adjustment amplitude according to actual spacing, effectively to adjust the travel speed of vehicle, makes actual spacing and setting
Dream car away from deviation minimize, will not only directly be determined according to the deviation amplitude of actual spacing and the dream car of setting away between
The adjustment amplitude of travel speed, therefore the situation for causing vehicle to jolt in the absence of the switching constantly between acceleration and deceleration, with
Stability, reliability and the comfortableness of car are improved.
Brief description of the drawings
Technical scheme in order to illustrate the embodiments of the present invention more clearly, makes required in being described below to embodiment
Accompanying drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the present invention, for
For those of ordinary skill in the art, on the premise of not paying creative work, other can also be obtained according to these accompanying drawings
Accompanying drawing.
Fig. 1 is a kind of structural representation of intelligence with vehicle control of the offer of the embodiment of the present invention one;
Fig. 2 is a kind of structural representation of intelligence with vehicle control of the offer of the embodiment of the present invention two;
Fig. 3 is the structural representation for the neutral net that the embodiment of the present invention two is provided;
Fig. 4 is the function curve diagram for the action neutral net that the embodiment of the present invention two is provided;
Fig. 5 is a kind of flow chart of intelligence with car control method of the offer of the embodiment of the present invention three;
Fig. 6 is a kind of flow chart of intelligence with car control method of the offer of the embodiment of the present invention four.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing to embodiment party of the present invention
Formula is described in further detail.
Embodiment one
The embodiments of the invention provide a kind of intelligence with vehicle control, situation of the low speed with car is particularly suitable for use in, referring to
Fig. 1, the intelligence includes with vehicle control:
Detection module 101, the actual spacing for detecting vehicle and front vehicles in real time, and according to actual spacing produce with
The corresponding enhancing signal of actual spacing, enhancing signal is used to representing deviation width of the dream car of actual spacing and setting away between
Degree;
Strengthen study module 102, for by the way of enhancing study, according to actual spacing and enhancing signal, determining car
Travel speed adjustment amplitude;
With car module 103, for according to adjustment amplitude, the travel speed of vehicle being adjusted, to change actual spacing.
It is readily apparent that, enhancing study is by an Autonomous Agent that can perceive environment (agent), autonomous learning selection energy
Reach the optimal action of its target.The process of specific autonomous learning is that agency makes action in its environment, and environment can be given instead
Feedback, acts on behalf of the feedback (successfully award, unsuccessfully give and punish) according to environment, and action is recognized and learnt, so that
Correct behavior is paid the utmost attention in follow-up action and avoids the behavior of mistake from occurring, so constantly study, may finally be determined
Go out optimal action.With reference to the present embodiment, detection module 101 detects environmental information (i.e. the actual spacing of vehicle and front vehicles),
Enhancing study module 102 first arbitrarily determines an action policy (the i.e. adjustment width of the travel speed of vehicle according to environmental information
Degree), acted (travel speed for adjusting vehicle) according to the action policy of determination with car module 103.Then detection module
101 detection environmental information and give according to the situation of change of environment again and feed back and (strengthen signal), strengthen study module 102
According to enhancing signal update action policy, acted with car module 103 according to the action policy after renewal.So constantly adjustment,
Until determining optimal action policy (i.e. actual spacing remain the dream car of setting away from).
Further, after being acted with car module 103, the actual spacing of vehicle and front vehicles can change, real
The deviation amplitude of border spacing and the dream car of setting away between may reduce, it is also possible to increase.If deviation amplitude reduces, pass through
The punishment that enhancing signal is given reduces therewith, and action of the enhancing study of study module 102 before is correct, and meeting is preferential later
Carried out using this action with car;If deviation amplitude increases, the punishment given by strengthening signal increases therewith, enhancing study
Action of the study of module 102 before is wrong, can avoid carrying out with car using this action later.Specifically, reality is worked as
When spacing is more than the following distance of setting, actual spacing is caused to be less than setting due to front vehicles Reduced Speed Now for the first time
During following distance, enhancing study module 102 can accordingly be punished by strengthening signal, will be kept away as far as possible when secondary
Exempt from identical situation occur, reduce jolting for vehicle, improve stability, reliability and comfortableness with car.
The embodiment of the present invention, according to actual spacing and enhancing signal, determines vehicle by using the mode of enhancing study
The adjustment amplitude of travel speed, enhancing signal, can for representing deviation amplitude of the dream car of actual spacing and setting away between
With during the adjustment of travel speed, using adaptive dynamic programming method, constantly by strengthening the deviation width that signal reflects
How degree, autonomous learning determines suitable adjustment amplitude according to actual spacing, effectively to adjust the travel speed of vehicle, makes reality
Spacing and the dream car of setting away from deviation minimize, will not only deviation according to actual spacing and the dream car of setting away between
Amplitude directly determines the adjustment amplitude of travel speed, therefore causes vehicle to run in the absence of constantly switching between acceleration and deceleration
Situation about winnowing with a dustpan, stability, reliability and comfortableness with car are improved.
Embodiment two
The embodiments of the invention provide a kind of intelligence with vehicle control, the present embodiment is the intelligence provided embodiment one
With the specific discussion of vehicle control, referring to Fig. 2, the intelligence includes with vehicle control:
Detection module 201, the actual spacing for detecting vehicle and front vehicles in real time, and according to actual spacing produce with
The corresponding enhancing signal of actual spacing, enhancing signal is used to representing actual spacing with the dream car that sets away between (such as 15m)
Deviation amplitude;
Strengthen study module 202, for by the way of enhancing study, according to actual spacing and enhancing signal, determining car
Travel speed adjustment amplitude;
With car module 203, for according to adjustment amplitude, the travel speed of vehicle being adjusted, to change actual spacing.
In the present embodiment, enhancing signal can be set according to actual conditions.General actual spacing and the reason of setting
Think that the deviation amplitude between spacing is smaller, enhancing signal is bigger.For example, when deviation amplitude is 0, enhancing signal is 0;With collision
During dangerous or loss front vehicles, enhancing signal is -1.
In actual applications, detection module 201 can include:
Spacing acquiring unit 201a, for detecting vehicle and the actual spacing of front vehicles in real time;
Signal generation unit 201b, for the functional relation according to setting, by the car of actual spacing and dream car away between
Away from deviation, enhancing signal is produced.
Specifically, the functional relation of setting can be linear function, the spacing deviation of actual spacing and dream car away between
Bigger, enhancing signal is smaller.
In actual applications, acquiring unit 201a can pass for camera, radar (such as millimetre-wave radar, laser radar)
Sensor, signal generation unit 201b can be single-chip microcomputer.
In a kind of implementation of the present embodiment, the enhancing study module 202 can include:
Neutral net 202a is acted, for according to actual spacing, producing the adjustment amplitude of the travel speed of vehicle;
Neutral net 202b is evaluated, for according to actual spacing, enhancing signal and adjustment amplitude, producing cost function,
Cost function is the approximate representation of enhancing signal;According to cost function, neutral net 202b neural network weight is evaluated in regulation,
To minimize the error of cost function and enhancing signal;
Action neutral net 202a is additionally operable to, what the evaluation neutral net 202b after being adjusted according to neural network weight was produced
Cost function, regulation action neutral net 202a neural network weight, obtains optimal adjustment amplitude, to minimize cost letter
Number and the error of desired value, desired value be actual spacing reach dream car away from when the cost function that produces.
It should be noted that action neutral net 202a neural network weight, evaluation neutral net 202b nerve net
Network weights initial value with random arrangement, can strengthen the process of the study of study module 202 mainly in configuration initial value to initial value
Constantly adjustment, to the last obtains optimal weights, now error reaches minimum.
Alternatively, action neutral net 202a can be used for,
Then to the neural network weight for acting neutral net 202a set the regulation of number of times using gradient descent method.
Alternatively, neutral net 202b is evaluated can be used for,
Then to the neural network weight for evaluating neutral net 202b set the regulation of number of times using gradient descent method.
It is to be appreciated that being then adjusted using gradient descent method, orderly regulation on the one hand can be realized, regulation is improved
Accuracy, on the other hand with Step wise approximation, can improve the efficiency of regulation.
In the specific implementation, neutral net 202b neural network weight is first evaluated according to setting number of times regulation, to reduce
The error of cost function and enhancing signal;Then the neural network weight of the evaluation neutral net 202b after updating is kept, according to
Number of times regulation action neutral net 202a neural network weight is set, by changing adjustment amplitude, to minimize cost function
With the error of desired value.And start to adjust the neural network weight ... for evaluating neutral net 202b so recycled back,
To the last action neutral net 202a neural network weight, evaluation neutral net 202b neural network weight tend to be steady
Fixed, the weights now obtained are optimal, and cost function reaches minimum, are hereafter adjusted according to the weights, quickly just
The ideal distance of setting can be reached.
It is to be appreciated that first neutral net 202b neural network weight is evaluated in regulation, cost function can be made more to approach
Strengthen signal, improve the accuracy evaluated;Action neutral net 202a neural network weight is adjusted again, you can regulation adjustment width
Degree (action neutral net 202a output valve and the input value of evaluation neutral net 202b neural network weight), and then adjust
Cost function, reduces the deviation of itself and desired value.
Specifically, evaluating neutral net 202b can be according to equation below (1) calculation error, and regulation is completed when error is 0:
ec(t)=α * J (t)-[J (t-1)-r (t)]; (1)
Wherein, ec(t) error of cost function and enhancing signal is represented, α is commutation factor, and 0 < α < 1, J (t) represents generation
Valency function, r (t) represents enhancing signal, and t represents the moment.
Acting neutral net 202a can be according to equation below (2) calculation error, and regulation is completed when error is 0:
ea(t)=J (t)-U (t); (2)
Wherein, ea(t) error of cost function and desired value is represented, J (t) represents cost function, and U (t) represents desired value,
T represents the moment.
In actual applications, (3) e (t) can be converted into E (t) it is adjusted as follows:
E (t)=(1/2) * [e (t)]2; (3)
Wherein, e (t) is ecOr e (t)a(t)。
It should be noted that it is to seek local derviation to weights to be then adjusted using gradient descent method, e (t) is converted into E (t),
Can be in order to calculating.
Specifically, evaluating neutral net 202b can be adjusted according to equation below (4)-(6):
wc(t+1)=wc(t)+Δwc(t); (4)
Wherein, wcRepresent evaluation neutral net 202b input value, △ wcRepresent that evaluating the middle of neutral net 202b ties
Really, lcRepresent evaluation neutral net 202b learning rate, EcEvaluation neutral net 202b error is represented, J represents cost letter
Number, t represents the moment.
Action neutral net 202a can be adjusted according to equation below (7)-(9):
wa(t+1)=wa(t)+Δwa(t); (7)
Wherein, waExpression action neutral net 202a input value, △ waKnot in the middle of expression action neutral net 202a
Really, laExpression action neutral net 202a learning rate, EaExpression action neutral net 202a error, J represents cost letter
Number, t represents the moment.
Preferably, action neutral net 202a and evaluation neutral net 202b can use Nonlinear Multi perceptron.
Multilayer perceptron includes one or more layers hidden layer, and each hidden layer includes some nodes, every layer each node respectively with phase
There are mapping relations in each node of adjacent bed.By taking one layer of hidden layer shown in Fig. 3 as an example, x1-xn is n input value, and y1-ym is
M node of hidden layer, z is output valve, and xi is mapped to node yj with weight w ij, and node yj calculates all input values
As a result it is mapped to weights vj in output valve.Wherein, n >=1 and n are integer, and m >=1 and m are integer, and 1≤i≤n and i are integer,
1≤j≤m and j are integer.With reference to action neutral net 202a, input value is actual spacing, and output valve is adjustment amplitude;With reference to
Neutral net 202b is evaluated, input value is actual spacing, enhancing signal, adjustment amplitude, and output valve is cost function.
It is readily apparent that, the level of neutral net is more, and relation is more complicated, the applicability and accuracy for handling event are higher,
Act neutral net 202a and evaluate neutral net 202b and use Nonlinear Multi perceptron, action neutral net can be improved
202a and the ability for evaluating neutral net 202b autonomous learnings, autonomous learning determine with car strategy closer to optimal, with car effect
Fruit is more preferably.
In actual applications, detection module 201 can also include:
Normalization unit 201c, for by actual spacing and enhancing signal normalization.
Alternatively, adjustment amplitude can be with symbolization sgn functions.
For example, the physical relationship of adjustment amplitude and actual spacing can be with as shown in figure 4, from fig. 4, it can be seen that adjustment amplitude
Span be [- 1,1].Wherein, adjustment amplitude is more than 0 and represents to need open out, and 1 expression throttle is added to maximum, 0 indicate without
Need open out or touch on the brake, adjustment amplitude is less than 0 expression needs and touched on the brake, and -1 expression brake is stepped on maximum.
That is, action neutral net 202a and evaluate neutral net 202b input value span for [- 1,
1].In the specific implementation, first each input value can be normalized, then input action neutral net 202a or evaluation again
Neutral net 202b.
Further, action neutral net 202a and evaluation neutral net 202b can use sigmoid functions.
It is to be appreciated that all numerical value are normalized, it is more convenient for calculating, also allows for making system be applied to all vehicles.
In actual applications, amplitude can be trampled according to accelerator and brake and the corresponding relation of operating range is specifically adjusted
Section, so that the present invention can be realized on various vehicles.
The embodiment of the present invention, according to actual spacing and enhancing signal, determines vehicle by using the mode of enhancing study
The adjustment amplitude of travel speed, enhancing signal, can for representing deviation amplitude of the dream car of actual spacing and setting away between
With during the adjustment of travel speed, using adaptive dynamic programming method, constantly by strengthening the deviation width that signal reflects
How degree, autonomous learning determines suitable adjustment amplitude according to actual spacing, effectively to adjust the travel speed of vehicle, makes reality
Spacing and the dream car of setting away from deviation minimize, will not only deviation according to actual spacing and the dream car of setting away between
Amplitude directly determines the adjustment amplitude of travel speed, therefore causes vehicle to run in the absence of constantly switching between acceleration and deceleration
Situation about winnowing with a dustpan, stability, reliability and comfortableness with car are improved.
Embodiment three
Referring to Fig. 5, the embodiments of the invention provide a kind of intelligence with car control method, it is adaptable to embodiment one or embodiment
Two intelligence provided are with vehicle control, and this method includes:
Step 301:Detection vehicle and the actual spacing of front vehicles, and produced and actual spacing according to actual spacing in real time
Corresponding enhancing signal.
In the present embodiment, enhancing signal is for representing deviation amplitude of the dream car of actual spacing and setting away between.
Step 302:By the way of enhancing study, according to actual spacing and enhancing signal, the travel speed of vehicle is determined
Adjustment amplitude.
Step 303:According to adjustment amplitude, the travel speed of vehicle is adjusted, to change actual spacing.
The embodiment of the present invention, according to actual spacing and enhancing signal, determines vehicle by using the mode of enhancing study
The adjustment amplitude of travel speed, enhancing signal, can for representing deviation amplitude of the dream car of actual spacing and setting away between
With during the adjustment of travel speed, using adaptive dynamic programming method, constantly by strengthening the deviation width that signal reflects
How degree, autonomous learning determines suitable adjustment amplitude according to actual spacing, effectively to adjust the travel speed of vehicle, makes reality
Spacing and the dream car of setting away from deviation minimize, will not only deviation according to actual spacing and the dream car of setting away between
Amplitude directly determines the adjustment amplitude of travel speed, therefore causes vehicle to run in the absence of constantly switching between acceleration and deceleration
Situation about winnowing with a dustpan, stability, reliability and comfortableness with car are improved.
Example IV
Referring to Fig. 6, the embodiments of the invention provide a kind of intelligence with car control method, it is adaptable to embodiment one or embodiment
Two intelligence provided are with vehicle control, and the present embodiment is the intelligent specific opinion with car control method provided embodiment three
State, this method includes:
Step 401:Detection vehicle and the actual spacing of front vehicles, and produced and actual spacing according to actual spacing in real time
Corresponding enhancing signal.
In the present embodiment, enhancing signal is for representing deviation amplitude of the dream car of actual spacing and setting away between.
Step 402:By the way of enhancing study, according to actual spacing and enhancing signal, the travel speed of vehicle is determined
Adjustment amplitude.
In a kind of implementation of embodiment, the step 402 can include:
Step 402a:Using action neutral net according to actual spacing, the adjustment amplitude of the travel speed of vehicle is produced.
Step 402b:Using neutral net is evaluated according to actual spacing, enhancing signal and adjustment amplitude, cost is produced
Function.
In the present embodiment, cost function is the approximate representation of enhancing signal.
Step 402c:According to cost function, the neural network weight of neutral net is evaluated in regulation, to minimize cost function
Error with strengthening signal.
Alternatively, step 402c can include:
Then to the neural network weight for acting neutral net set the regulation of number of times using gradient descent method.
Step 402d:The cost function that evaluation neutral net after being adjusted according to neural network weight is produced, regulation action
The neural network weight of neutral net, obtains optimal adjustment amplitude, to minimize the error of cost function and desired value.
In the present embodiment, desired value be actual spacing reach dream car away from when the cost function that produces.
Alternatively, step 402d can include:
Then to the neural network weight for evaluating neutral net set the regulation of number of times using gradient descent method.
Preferably, action neutral net and evaluation neutral net can use Nonlinear Multi perceptron.
Step 403:According to adjustment amplitude, the travel speed of vehicle is adjusted, to change actual spacing.
Alternatively, adjustment amplitude can be with symbolization function.
The embodiment of the present invention, according to actual spacing and enhancing signal, determines vehicle by using the mode of enhancing study
The adjustment amplitude of travel speed, enhancing signal, can for representing deviation amplitude of the dream car of actual spacing and setting away between
With during the adjustment of travel speed, using adaptive dynamic programming method, constantly by strengthening the deviation width that signal reflects
How degree, autonomous learning determines suitable adjustment amplitude according to actual spacing, effectively to adjust the travel speed of vehicle, makes reality
Spacing and the dream car of setting away from deviation minimize, will not only deviation according to actual spacing and the dream car of setting away between
Amplitude directly determines the adjustment amplitude of travel speed, therefore causes vehicle to run in the absence of constantly switching between acceleration and deceleration
Situation about winnowing with a dustpan, stability, reliability and comfortableness with car are improved.
It should be noted that:Above-described embodiment provide intelligence with vehicle control control intelligence with car when, only more than
The division progress of each functional module is stated for example, in practical application, as needed can distribute above-mentioned functions by difference
Functional module complete, i.e., the internal structure of system is divided into different functional modules, with complete it is described above whole or
Person's partial function.In addition, the intelligence that above-described embodiment is provided belongs to vehicle control with intelligence with car control method embodiment
Same design, it implements process and refers to embodiment of the method, repeats no more here.
The embodiments of the present invention are for illustration only, and the quality of embodiment is not represented.
One of ordinary skill in the art will appreciate that realizing that all or part of step of above-described embodiment can be by hardware
To complete, the hardware of correlation can also be instructed to complete by program, described program can be stored in a kind of computer-readable
In storage medium, storage medium mentioned above can be read-only storage, disk or CD etc..
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc. should be included in the scope of the protection.
Claims (8)
1. a kind of intelligence is with vehicle control, it is characterised in that the intelligence includes with vehicle control:
Detection module, is produced and institute for detecting vehicle and the actual spacing of front vehicles in real time, and according to the actual spacing
The corresponding enhancing signal of actual spacing is stated, the enhancing signal is for representing the dream car of the actual spacing and setting away between
Deviation amplitude;
Strengthen study module, for by the way of enhancing study, according to the actual spacing and the enhancing signal, determining institute
State the adjustment amplitude of the travel speed of vehicle;
With car module, for according to the adjustment amplitude, the travel speed of the vehicle being adjusted, to change the actual spacing;
The enhancing study module, including:
Neutral net is acted, for according to the actual spacing, producing the adjustment amplitude of the travel speed of the vehicle;
Neutral net is evaluated, for according to the actual spacing, the enhancing signal and the adjustment amplitude, producing cost
Function, the cost function strengthens the approximate representation of signal to be described;According to the cost function, the evaluation nerve net is adjusted
The neural network weight of network, to minimize the error of the cost function and the enhancing signal;
The action neutral net is additionally operable to, described in the evaluation neutral net generation after being adjusted according to neural network weight
Cost function, adjusts the neural network weight of the action neutral net, optimal adjustment amplitude is obtained, to minimize the generation
The error of valency function and desired value, the desired value be the actual spacing reach the dream car away from when the cost that produces
Function.
2. intelligence according to claim 1 is with vehicle control, it is characterised in that the action neutral net and institute's commentary
Valency neutral net uses Nonlinear Multi perceptron.
3. intelligence according to claim 1 is with vehicle control, it is characterised in that the action neutral net is used for,
Then set the regulation of number of times to the neural network weight of the action neutral net using gradient descent method;
The evaluation neutral net is used for,
Then set the regulation of number of times to the neural network weight of the evaluation neutral net using gradient descent method.
4. the intelligence according to claim any one of 1-3 is with vehicle control, it is characterised in that the adjustment amplitude is used
Sign function.
5. a kind of intelligence is with car control method, it is characterised in that the intelligence includes with car control method:
Detection vehicle and the actual spacing of front vehicles, and in real time corresponding with the actual spacing according to the actual spacing generation
Enhancing signal, deviation amplitude of the dream car that the enhancing signal is used to representing the actual spacing and setting away between;
By the way of enhancing study, according to the actual spacing and the enhancing signal, the travel speed of the vehicle is determined
Adjustment amplitude;
According to the adjustment amplitude, the travel speed of the vehicle is adjusted, to change the actual spacing;
It is described that the adjustment amplitude of the travel speed of the vehicle is determined according to the actual spacing and the enhancing signal, including:
Using action neutral net according to the actual spacing, the adjustment amplitude of the travel speed of the vehicle is produced;
Using neutral net is evaluated according to the actual spacing, the enhancing signal and the adjustment amplitude, cost letter is produced
Number, the cost function strengthens the approximate representation of signal to be described;
According to the cost function, the neural network weight for evaluating neutral net is adjusted, to minimize the cost function
With the error of the enhancing signal;
The cost function that the evaluation neutral net after being adjusted according to neural network weight is produced, adjusts the action god
Neural network weight through network, obtains optimal adjustment amplitude, to minimize the error of the cost function and desired value, institute
State desired value for the actual spacing reach the dream car away from when the cost function that produces.
6. intelligence according to claim 5 is with car control method, it is characterised in that the action neutral net and institute's commentary
Valency neutral net uses Nonlinear Multi perceptron.
7. intelligence according to claim 5 is with car control method, it is characterised in that described according to the cost function, adjust
The section neural network weight for evaluating neutral net, including:
Then set the regulation of number of times to the neural network weight of the action neutral net using gradient descent method;
It is described adjusted according to neural network weight after the cost function that produces of the evaluation neutral net, adjust described dynamic
Make the neural network weight of neutral net, including:
Then set the regulation of number of times to the neural network weight of the evaluation neutral net using gradient descent method.
8. the intelligence according to claim any one of 5-7 is with car control method, it is characterised in that the adjustment amplitude is used
Sign function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510489170.5A CN105059213B (en) | 2015-08-11 | 2015-08-11 | A kind of intelligence is with vehicle control and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510489170.5A CN105059213B (en) | 2015-08-11 | 2015-08-11 | A kind of intelligence is with vehicle control and method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105059213A CN105059213A (en) | 2015-11-18 |
CN105059213B true CN105059213B (en) | 2017-08-04 |
Family
ID=54488819
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510489170.5A Active CN105059213B (en) | 2015-08-11 | 2015-08-11 | A kind of intelligence is with vehicle control and method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105059213B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105835854B (en) * | 2016-03-17 | 2018-11-16 | 奇瑞汽车股份有限公司 | A kind of emergency braking control system and its control method |
CN107065561B (en) * | 2017-05-16 | 2019-11-22 | 清华大学 | The machine learning control method of double-wheel self-balancing vehicle |
CN109686086B (en) * | 2018-12-24 | 2020-08-07 | 东软集团(北京)有限公司 | Method and device for training fuzzy control network and generating intersection suggested speed |
CN109849911B (en) * | 2019-02-19 | 2020-10-09 | 百度在线网络技术(北京)有限公司 | Car following method, car following device and computer readable storage medium |
CN110194156B (en) * | 2019-06-21 | 2020-11-10 | 厦门大学 | Intelligent network-connected hybrid electric vehicle active collision avoidance reinforcement learning control system and method |
CN112100855A (en) * | 2020-09-17 | 2020-12-18 | 北京智能车联产业创新中心有限公司 | Vehicle following capability evaluation method and device, electronic device and storage medium |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08127268A (en) * | 1994-10-28 | 1996-05-21 | Isuzu Motors Ltd | Vehicle running controller |
CN101417655A (en) * | 2008-10-14 | 2009-04-29 | 清华大学 | Vehicle multi-objective coordinated self-adapting cruise control method |
JP2011095834A (en) * | 2009-10-27 | 2011-05-12 | Denso Corp | Device and program for controlling vehicle |
CN103129556A (en) * | 2011-11-22 | 2013-06-05 | 罗伯特·博世有限公司 | Driving assistance system |
KR20130059702A (en) * | 2011-11-29 | 2013-06-07 | 현대자동차주식회사 | Apparatus and method for controlling obstacle sensing zone |
CN104024021A (en) * | 2012-01-02 | 2014-09-03 | 沃尔沃拉斯特瓦格纳公司 | Method and system for controlling a driving distance |
CN104554264A (en) * | 2014-11-28 | 2015-04-29 | 温州大学 | Method and system for self-adaptively on-line adjusting vehicle speed |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009149173A (en) * | 2007-12-19 | 2009-07-09 | Mitsubishi Fuso Truck & Bus Corp | Auto-cruise device |
-
2015
- 2015-08-11 CN CN201510489170.5A patent/CN105059213B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08127268A (en) * | 1994-10-28 | 1996-05-21 | Isuzu Motors Ltd | Vehicle running controller |
CN101417655A (en) * | 2008-10-14 | 2009-04-29 | 清华大学 | Vehicle multi-objective coordinated self-adapting cruise control method |
JP2011095834A (en) * | 2009-10-27 | 2011-05-12 | Denso Corp | Device and program for controlling vehicle |
CN103129556A (en) * | 2011-11-22 | 2013-06-05 | 罗伯特·博世有限公司 | Driving assistance system |
KR20130059702A (en) * | 2011-11-29 | 2013-06-07 | 현대자동차주식회사 | Apparatus and method for controlling obstacle sensing zone |
CN104024021A (en) * | 2012-01-02 | 2014-09-03 | 沃尔沃拉斯特瓦格纳公司 | Method and system for controlling a driving distance |
CN104554264A (en) * | 2014-11-28 | 2015-04-29 | 温州大学 | Method and system for self-adaptively on-line adjusting vehicle speed |
Also Published As
Publication number | Publication date |
---|---|
CN105059213A (en) | 2015-11-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105059213B (en) | A kind of intelligence is with vehicle control and method | |
CN105059288B (en) | A kind of system for lane-keeping control and method | |
CN111898211B (en) | Intelligent vehicle speed decision method based on deep reinforcement learning and simulation method thereof | |
CN105109488B (en) | A kind of intelligence is with car system and method | |
Zhao et al. | Chassis integrated control for active suspension, active front steering and direct yaw moment systems using hierarchical strategy | |
CN105035085B (en) | Automatically with car method and device | |
CN102109821B (en) | System and method for controlling adaptive cruise of vehicles | |
CN103324085B (en) | Based on the method for optimally controlling of supervised intensified learning | |
CN110304045A (en) | Intelligent driving transverse direction lane-change decision-making technique, system and device | |
KR102166811B1 (en) | Method and Apparatus for Controlling of Autonomous Vehicle using Deep Reinforcement Learning and Driver Assistance System | |
CN103693040A (en) | Vehicle collision avoidance system based on double-mode cooperation | |
CN108694841A (en) | A kind of intelligent vehicle passage crossroads traffic light method based on V2X technologies | |
CN114013443A (en) | Automatic driving vehicle lane change decision control method based on hierarchical reinforcement learning | |
CN105835854B (en) | A kind of emergency braking control system and its control method | |
Zhang et al. | Adaptive shared control strategy for lane changing assistance system via multi-mode switching | |
Selvaraj et al. | An ML-aided reinforcement learning approach for challenging vehicle maneuvers | |
Xu et al. | Hierarchical speed control for autonomous electric vehicle through deep reinforcement learning and robust control | |
Zhao et al. | Coordinated throttle and brake fuzzy controller design for vehicle following | |
CN105109485A (en) | Driving method and system | |
Zhao et al. | Supervised adaptive dynamic programming based adaptive cruise control | |
Xu et al. | Modeling Lateral Control Behaviors of Distracted Drivers for Haptic-Shared Steering System | |
Nageshrao et al. | Robust ai driving strategy for autonomous vehicles | |
Guo et al. | Adaptive Lane-Departure Prediction Method with Support Vector Machine and Gated Recurrent Unit Models | |
CN104925059B (en) | Automatically with car method and device | |
Hailemichael et al. | Safe reinforcement learning for an energy-efficient driver assistance system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20220218 Address after: 241006 Anshan South Road, Wuhu Economic and Technological Development Zone, Anhui Province Patentee after: Wuhu Sambalion auto technology Co.,Ltd. Address before: 241006 Changchun Road, Wuhu economic and Technological Development Zone, Wuhu, Anhui, 8 Patentee before: CHERY AUTOMOBILE Co.,Ltd. |
|
TR01 | Transfer of patent right |