CN114237275A - Multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery - Google Patents
Multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery Download PDFInfo
- Publication number
- CN114237275A CN114237275A CN202111158497.6A CN202111158497A CN114237275A CN 114237275 A CN114237275 A CN 114237275A CN 202111158497 A CN202111158497 A CN 202111158497A CN 114237275 A CN114237275 A CN 114237275A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- target
- probability
- search
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000005381 potential energy Methods 0.000 claims abstract description 23
- 230000004927 fusion Effects 0.000 claims abstract description 14
- 230000007246 mechanism Effects 0.000 claims abstract description 13
- 238000001514 detection method Methods 0.000 claims description 28
- 230000006870 function Effects 0.000 claims description 18
- 238000003384 imaging method Methods 0.000 claims description 12
- 230000009471 action Effects 0.000 claims description 10
- 230000008447 perception Effects 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 8
- 230000003993 interaction Effects 0.000 claims description 7
- 230000007613 environmental effect Effects 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 5
- 238000004891 communication Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 239000004576 sand Substances 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000001149 cognitive effect Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 238000012634 optical imaging Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000007500 overflow downdraw method Methods 0.000 abstract description 6
- 230000000694 effects Effects 0.000 abstract description 3
- 238000004088 simulation Methods 0.000 description 3
- 239000003795 chemical substances by application Substances 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 231100001261 hazardous Toxicity 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010348 incorporation Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/10—Simultaneous control of position or course in three dimensions
- G05D1/101—Simultaneous control of position or course in three dimensions specially adapted for aircraft
- G05D1/106—Change initiated in response to external conditions, e.g. avoidance of elevated terrain or of no-fly zones
Landscapes
- Engineering & Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
A multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery belongs to the technical field of multi-unmanned aerial vehicle collaborative search. Firstly, a plurality of unmanned aerial vehicles cooperatively search and establish an environment model and initialization parameters; secondly, setting task load parameters of the unmanned aerial vehicles to complete a flight decision potential energy game of the multiple unmanned aerial vehicles; and finally updating the target probability graph through a perception-locking-discovery mechanism. The invention updates the environment probability graph by searching the target, thereby reducing the uncertainty of the whole environment; a perception-locking-discovery search mechanism is provided, the discovery probability of a target is increased, a higher-value probability graph fusion method is provided, the limitation of self information is overcome, and meanwhile, the effectiveness and the accuracy of information fusion are improved; through the potential energy game method, the autonomous decision control of multiple unmanned aerial vehicles is facilitated, and the overall searching efficiency is ensured. The method is particularly suitable for the condition that the unmanned aerial vehicle task load performance is low, the lifting effect is obvious, and the method has a good application prospect.
Description
Technical Field
The invention relates to the technical field of unmanned aerial vehicle collaborative search, in particular to a multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery.
Background
The rapid development of Unmanned Aerial Vehicles (UAVs) has facilitated their incorporation into many areas. Unmanned aerial vehicle is used for carrying out the complex task in the hazardous environment owing to have advantages such as with low costs, mobility is strong, avoid casualties. However, due to the low efficiency and poor robustness of single drone in task execution, the use of multiple drones for cooperative task execution is receiving more and more attention in military and civil applications, and cooperative search is a major application aspect of multiple drones equipped with reconnaissance loads (such as cameras, radars and sonars).
Many search and surveillance tasks involve measuring and exploring unknown areas, such as object search, environmental monitoring, mapping, etc., which can achieve higher efficiency and significantly reduce risk cost by means of collaborative decision-making through information interaction between multiple drones. Collaborative search involves the design of distributed algorithms, i.e., global optimization goals of the entire system are achieved through local information. In the multi-unmanned aerial vehicle collaborative search problem, the following main technical problems need to be considered: (1) environment representation and updating, how to represent the presence and uncertainty of the target in the environment; (2) the performance of the task load, how to seek the balance between the detection radius and the detection probability, so that the unmanned aerial vehicle can more effectively search the target in the task area; (3) search path planning, how to design a cooperative control method, enables the unmanned aerial vehicle to move in a manner of maximizing the possibility of finding a target or minimizing environmental uncertainty. The purpose of the collaborative search is to control a plurality of unmanned aerial vehicles to search for unknown ground targets scattered in a task area, simultaneously reduce the uncertainty of the environment to the maximum extent and minimize the search time.
The game theory is a powerful tool for solving the multi-agent decision problem, mainly solves the problems of interactive communication, cooperative cooperation, conflict elimination and the like, is widely applied to the field of distributed computing, and becomes a research hotspot of multi-unmanned aerial vehicle cooperative search.
Disclosure of Invention
The invention aims to provide a collaborative search method, so that an unmanned aerial vehicle can search more effectively and reduce the uncertainty of the environment according to the conversion of a sensing-locking-discovering stage when executing a search task, and the collaborative search of multiple unmanned aerial vehicles can achieve the optimal overall efficiency through methods such as large-value probability graph fusion and potential energy game.
The multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery comprises the following steps:
step 1: modeling the multi-unmanned aerial vehicle collaborative search environment and the flight state;
step 2: unmanned aerial vehicle searching task load modeling;
and step 3: establishing a single unmanned aerial vehicle target probability map updating mechanism;
and 4, step 4: fusing a multi-unmanned aerial vehicle collaborative search target probability graph;
and 5: and (4) solving Nash equilibrium through the potential energy game, wherein the Nash equilibrium point corresponds to a global or local optimal solution of the multi-unmanned aerial vehicle path decision in the current state.
Preferably, the modeling of the multi-unmanned aerial vehicle collaborative search environment and the flight state in step 1 of the invention specifically comprises the following steps:
searching unknown environments
Equally divide the environment into Lx×LyThe units are the same in size, and each unit g is marked by (x, y);
mission drone tagging
The nth unmanned plane position (x)n，yn，hn) Identification of, wherein
hn∈[hmin，hmax]；hmin，hmaxRespectively the minimum and maximum flight heights of the unmanned aerial vehicle;
object marking requiring search
(x) for the t-th target positiont，yt) Identifying that the tth target is positioned on the ground with the height of 0 and represents that the target exists in the unit as long as the target position falls in the unit; the objective presence of an object is denoted by ω, 1 represents the object in the cell, 0 represents the object not in the cell; unmanned aerial vehicle detection results are expressed by xi, 1 represents that an object is found in a unit, and 0 represents that no object is found in the unit;
threat zone tagging
(x) for mth threat zonem，ym，Rm) Marking, the threat range is hemispherical, the sphere center of the threat range is positioned on the ground with the height of 0, and the threat radius is Rm；
All the positions of the points satisfy xn，xt，xm∈{1，2，…，LxAnd yn，yt，ym∈{1，2，…，Ly}。
{θn(t)，ηn(t) represents the flight state of the nth unmanned aerial vehicle at the time t, respectively represents the change of a course corner and a course height, is constrained by the flight performance of the unmanned aerial vehicle, and turns 45 degrees left, straight or turns 45 degrees right on the basis of the course of the unmanned aerial vehicle at the time t +1, and the maneuverability constraint conditions required to be met are as follows:
|ηn(t+1)-ηn(t)|≤ηmax
wherein eta ismaxRepresenting the maximum change value of the flight height of the unmanned aerial vehicle;
the flight conditions of the drone also satisfy the following constraints:
dij(t)≥dsafe(i，j＝1，2，…，N，i≠j)
dim(t)≥Rm(i＝1，2，…，N；m＝1，2，…，M)
wherein d isijThe distance between the ith unmanned aerial vehicle and the jth unmanned aerial vehicle is represented; dsafeRepresents a minimum safe distance between drones; dimRepresenting the distance between the ith drone and the mth threat.
Preferably, the unmanned aerial vehicle search task load modeling in step 2 of the present invention specifically includes the following steps:
the photoelectric load detection model describes the detection and discovery relation of the unmanned aerial vehicle to the search target;
general probability of detection PdIs represented by PdP (ξ ═ 1| ω ═ 1); false alarm probability PfIs represented by PfP (ξ ═ 1| ω ═ 0); the photoelectric load is vertically and downwards fixedly installed, and the imaging size delta of the target on the imaging surface of the airborne photoelectric load is known from the optical imaging principle as follows:
in the formula, HsFor unmanned aerial vehicle detection height, f is photoelectric load focal length, DcIs the characteristic size of the target; height of detection HsAnd a detection radius R5The relationship between them is:
in the formula, beta5The photoelectric load field angle; calculating the number of line pairs N of the critical dimension of the target covered on the photoelectric load imaging target surface as follows:
where b is the imaging size of the target surface of the photoelectric load, N5The number of scanning lines is the number of photoelectric load; for the identification of a specific target, an empirical rule of a required target resolution is established, empirical data required for finding, orienting, identifying and confirming the specific target in an image are given by using a Johnson criterion, and a calculation formula of a target transfer probability function can be reversely deduced according to a data table:
preferably, the single drone target probability map updating mechanism of step 3 of the present invention specifically includes the following steps:
the cooperative search of the multiple unmanned aerial vehicles is to execute tasks through perception of the unmanned aerial vehicles on unknown environments, information interaction between the unmanned aerial vehicles and cooperative decision; during the whole task, the unmanned aerial vehicle makes a decision according to the task load and the information of the adjacent unmanned aerial vehicle, so that the search task is cooperatively executed to realize the optimal configuration; the collaborative search includes the following three parts: task load observation, information fusion and cooperative motion; before searching, each unmanned aerial vehicle associates the pre-known environmental information with a probability map, and then the unmanned aerial vehicle moves to a position with high environmental uncertainty according to an algorithm to ensure the probability of target searching; the uncertainty of a corresponding area is reduced through task load observation, and in order to further improve the searching efficiency, the unmanned aerial vehicle carries out information fusion through communication with a neighbor, so that the unmanned aerial vehicle is guided to follow-up cooperative motion; the whole process then loops until the probability distribution over all targets or the whole task space is searched for a threshold.
Preferably, the multi-unmanned aerial vehicle collaborative search target probability map fusion in step 4 of the present invention specifically includes the following steps:
the unmanned aerial vehicle updates the target probability map of the unmanned aerial vehicle through information interaction, and combines the information of the unmanned aerial vehicle and the acquired information to fuse the target probability map; the target probability map is represented as grid-based probability cells, wherein each cell corresponds to a discrete search area having an associated target presence probability; each unit stores some useful information including the probability of the existence of the target, the uncertainty of the environment and the situation; combining all the units to obtain a cognitive information graph for searching;
Ftargetindicating whether the unmanned plane is in a locked state, when FtargetWhen the value is 0, the unmanned aerial vehicle is not allocated with a locking target, and the unmanned aerial vehicle flies according to the proposed algorithm; when F is presenttargetWhen the target is 1, the unmanned aerial vehicle is assigned to lock the target, the unmanned aerial vehicle starts to move to the target, and meanwhile, the flying height is reduced to improve the exploration probability of the load and ensure the identification rate of target search; plockedAnd PtargetRespectively representing a locking probability threshold and an object probability threshold when the unit object existence probability is higher than PlockedThen, the unit is listed in a target locking sequence to allocate the corresponding unmanned aerial vehicle to perform more accurate search; when the unit object existence probability is higher than PtargetThen, the unit is determined to exist at the target point;
Hthresholdswitching threshold representing the sensing and discovery phases when flight altitude is above HthresholdWhen in the sensing stage; when the flying height is lower than HthresholdWhen in the discovery phase; because the unmanned aerial vehicle is mostly in the higher condition of flight in perception phase, the P of load this momentdAnd PfAnd (3) carrying out probability map updating by a perception smoothing method after large change occurs:
where ρ represents a perceptual coefficient, and is generally taken to be 1-Pd；
The discovery phase is updated by bayesian consistency estimation.
Reducing the information entropy as an optimization target, namely reducing the uncertainty of the information in the unit;
S(t)＝-P(t)log2P(t)-(1-P(t)log21-P(t))。
preferably, the potential energy game in the step 5 of the invention solves nash equilibrium, and the nash equilibrium point corresponds to the global or local optimal solution of the multi-unmanned aerial vehicle path decision in the current state, and the method specifically comprises the following steps:
marking a multi-unmanned plane game model as
Wherein
The method comprises the steps that a set of game participants, namely a set of task unmanned aerial vehicles is obtained;
the action set of the nth unmanned aerial vehicle is obtained; u. ofnIs the utility function of the nth unmanned plane;
if action policy set
Wherein
And
such that the utility function satisfies:
then a*Belonging to the game model
Pure strategy Nash equilibrium point of (1), wherein a-nAn action policy representing all participants except n;
if there is an accurate potential energy function phi such that
Satisfies the following conditions:
u(an′，a-n)-u(an，a-n)＝φ(an′，a-n)-φ(an，a-n)
the game is referred to as an accurate potential energy game.
The invention provides a multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery. The environment probability graph is updated through searching the target, so that the uncertainty of the whole environment is reduced; a perception-locking-discovery search mechanism is provided, and compared with the traditional search method, the discovery probability of the target is increased; a higher-value probability map fusion method is provided, so that the information fusion effectiveness and accuracy are improved while the limitation of self information is overcome; through the potential energy game method, the autonomous decision control of multiple unmanned aerial vehicles is facilitated, and the overall searching efficiency is ensured. In addition, the method is very simple to implement, has an obvious improvement effect especially under the condition that the task load performance of the unmanned aerial vehicle is low, and has a good application prospect.
Drawings
Fig. 1 is a flowchart of a collaborative search procedure of multiple drones.
Fig. 2 is a scene diagram of collaborative search by multiple drones.
FIG. 3 is a diagram of task load field of view range versus detection probability.
FIG. 4 impact of the lock-in phase on the average target presence probability.
Figure 5 influence of the perception-discovery phase on the average target presence probability.
FIG. 6 impact of different fusion methods on the average object presence probability.
FIG. 7 impact of different algorithms on the average number of search targets.
Detailed Description
In order to realize more effective search and reduce the uncertainty of the environment, the invention adopts the following technical scheme that the cooperative search of multiple unmanned aerial vehicles achieves the optimal overall efficiency by methods such as higher-value probability graph fusion and potential energy game, and the like:
the multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery comprises the following steps:
(1) multi-unmanned aerial vehicle collaborative search environment and flight state modeling
Searching unknown environments
Equally divide the environment into Lx×LyThe units are of the same size, and each unit g is identified by (x, y).
Mission drone tagging
The nth unmanned plane position (x)n，yn，hn) Identification of, wherein
hn∈[hmin，hmax]；hmin，hmaxRespectively, unmanned aerial vehicle minimum and maximum flying height.
Object marking requiring search
(x) for the t-th target positiont，yt) A flag, which is located on the ground with a height of 0, represents that the target is present for the cell as long as the target location falls within the cell. The objective presence of an object is denoted by ω, with 1 representing the object in the cell and 0 representing the object not in the cell. The drone detection result is represented by ξ, 1 represents that an object is found in a cell, and 0 represents that no object is found in a cell.
Threat zone tagging
(x) for mth threat zonem，ym，Rm) Marking, the threat range is hemispherical, the sphere center of the threat range is positioned on the ground with the height of 0, and the threat radius is Rm。
All the positions of the points satisfy xn，xt，xm∈{1，2，…，LxAnd yn，yt，ym∈{1，2，…，Ly}。
{θn(t)，ηn(t) represents the flight state of the nth unmanned aerial vehicle at the time t, the flight states respectively represent the change of a course corner and the change of the height, and are constrained by the flight performance of the unmanned aerial vehicle, the course of the unmanned aerial vehicle at the time t +1 rotates 45 degrees left, moves straight or rotates 45 degrees right on the basis of the course at the time t, and the maneuverability constraint conditions required to be met are as follows:
|ηn(t+1)-ηn(t)|≤ηmax
wherein eta ismaxRepresenting the maximum variation value of the flying height of the unmanned aerial vehicle. For the collaborative search problem of multiple unmanned aerial vehicles, the unmanned aerial vehicles are considered to collide with each other, the unmanned aerial vehicles and the threat objects are also considered to keep a safe distance, and therefore the flight conditions of the unmanned aerial vehicles still meet the following constraints:
dij(t)≥dsafe(i，j＝1，2，…，N；i≠j)
dim(t)≥Rm(i＝1，2，…，N；m＝1，2，…，M)
wherein d isijThe distance between the ith unmanned aerial vehicle and the jth unmanned aerial vehicle is represented; dsafe represents the minimum safe distance between drones; dimRepresenting the distance between the ith drone and the mth threat.
(2) Unmanned aerial vehicle search task load modeling
The photoelectric load detection model describes the detection and discovery relation of the unmanned aerial vehicle on the search target. General probability of detection PdIs represented by PdP (ξ ═ 1| ω ═ 1); false alarm probability PfIs represented by PfP (ξ ═ 1| ω ═ 0). Considering the photoelectric load fixedly installed vertically downwards, the imaging size δ of the target on the imaging plane of the onboard photoelectric load is as follows from the optical imaging principle:
in the formula, HsFor unmanned aerial vehicle detection height, f is photoelectric load focal length, DcIs the characteristic dimension of the target. Height of detection HsAnd a detection radius RsThe relationship between them is:
in the formula, betasThe photoelectric load field angle. Therefore, the line pair number N of the target covering the critical dimension of the target on the photoelectric load imaging target surface can be calculated as follows:
where b is the imaging size of the target surface of the photoelectric load, NsThe number of line scans of the photoelectric load. For the identification of a specific target, an empirical rule of a required target resolution is established at present, the most common is the Johnson criterion, which gives empirical data required for the discovery, orientation, identification and confirmation of the specific target in an image, and a calculation formula of a target transfer probability function can be reversely deduced according to a data table.
(3) Single unmanned aerial vehicle target probability map updating mechanism
The key of the collaborative search of the multiple unmanned aerial vehicles is to execute tasks through perception of the unmanned aerial vehicles on unknown environments, information interaction between the unmanned aerial vehicles and collaborative decision. During the whole task, the unmanned aerial vehicle can make own decision according to the task load and the information of the adjacent unmanned aerial vehicle, so that the search task is executed cooperatively to realize the optimal configuration. The collaborative search problem involves the following three components: task load observation, information fusion and cooperative motion. Before searching is started, each unmanned aerial vehicle associates the previously known environment information with the probability map, and then the unmanned aerial vehicle moves to a position with high environment uncertainty according to an algorithm to ensure the probability of target searching. The uncertainty of the corresponding area is reduced through task load observation, and in order to further improve the searching efficiency, the unmanned aerial vehicle carries out information fusion through communication with neighbors, so that the unmanned aerial vehicle is guided to follow-up cooperative motion. The whole process then loops until the probability distribution over all targets or the whole task space is searched for a threshold. In the process, the invention ensures that the overall searching efficiency reaches the optimum through modes of a perception-locking-discovery mechanism, a potential energy game and the like.
(4) Multi-unmanned aerial vehicle collaborative search target probability map fusion
The unmanned aerial vehicle can update the probability map of the unmanned aerial vehicle through information interaction, and the probability map is fused by combining the information of the unmanned aerial vehicle and the acquired information. The probability map updates of the current study are mainly divided into three types: one is timestamp update, the other is weighted average update, and the third is an occupancy grid update. Although the timestamp updating can ensure that each unmanned aerial vehicle makes a decision according to the latest probability map, when the detection time difference of multiple unmanned aerial vehicles in the same unit is not large, the detection time difference is easily influenced by false alarm and missed alarm of subsequent unmanned aerial vehicles; although the error can be reduced by the weighted average method, the target probability is slowly increased and can reach the target probability threshold value through multiple observations; occupancy grid updates can make probability maps converge quickly, but tend to fall into cases where they cannot be updated, typically in conjunction with a weighted average approach. The invention provides a fusion method of a probability map with a larger value by combining the advantages and disadvantages of the method and the search characteristic of a perception-locking-discovery mechanism.
(5) Potential energy game solving Nash equilibrium
The potential energy game plays a prominent role in cooperative control of the distributed multi-agent system, and ensures that the local utility of each participant is consistent with the global target. Modeling the multi-unmanned aerial vehicle search problem into a mutual profit game model and proving that the game is an accurate potential energy game, wherein the Nash equilibrium point of the game corresponds to the global or local optimal solution of the multi-unmanned aerial vehicle path decision in the current state.
Marking a multi-unmanned plane game model as
Wherein
The method comprises the steps that a set of game participants, namely a set of task unmanned aerial vehicles is obtained;
the action set of the nth unmanned aerial vehicle is obtained; u. ofnThe utility function of the nth drone.
Defining: if action policy set
(wherein
And
) Such that the utility function satisfies:
then a*Belonging to the game model
Pure strategy Nash equilibrium point of (1), wherein a-nRepresenting the action policy of all participants except n.
Defining: if there is an accurate potential energy function phi such that
Satisfies the following conditions:
u(an′，a-n)-u(an，a-n)＝φ(an′，a-n)-φ(an，a-n)
the game is referred to as an accurate potential energy game.
As shown in figure 1, perception-locking-discovery-based multi-unmanned aerial vehicle game collaborative search method
The method comprises the following steps: multi-unmanned aerial vehicle collaborative search establishment environment model and initialization parameters
According to the modeling mode of the invention, 4 unmanned aerial vehicles are adopted to search 10 target points in 100 x 100 unknown areas, wherein 10 threat areas limit the unmanned aerial vehicles to fly, and fig. 2 shows a multi-unmanned aerial vehicle collaborative search scene graph. The minimum flying height of the unmanned aerial vehicle is 400 meters, and the maximum flying height is 1000 meters.
Step two: unmanned aerial vehicle task load parameter setting
Taking the number N of scanning lines of the load line of the airborne photoelectric tasks1080, angle of view βs80 °, the minimum dimensional directional resolution N of the found targetminThe relationship between the field range and the detection probability is shown in fig. 3. In order to better embody the effect of the collaborative search of the invention, the characteristic dimension D of the target is selected in the simulation c1m, so as to deduce the view field range R of the unmanned plane at different heightssAnd probability of detection Pd。
Step three: multi-unmanned aerial vehicle flight decision potential energy game
Designing a utility function in a mode of average sharing rules, marginal contribution, a Shapril value and the like to ensure that the game model is an accurate potential energy game, wherein the marginal contribution rules are that the utility function is derived through the potential energy function:
un(a)＝φ(a)-φ(a-n)
un(an′，a-n)-un(an，a-n)
＝φ(an′，a-n)-φ(a-n)-(φ(an，a-n)-φ(a-n)
＝φ(an′，a-n)-φ(an，a-n)
designing the potential energy function of each participant as the information entropy function sum in the unmanned aerial vehicle search range:
the utility function of the nth unmanned plane is designed as marginal contribution to global utility
Therefore, the Nash equilibrium points can be solved, and the action strategy set a of each unmanned aerial vehicle is obtained*。
Step four: perception-locking-discovery mechanism target probability map update
The target probability map is represented as grid-based probability cells, where each cell corresponds to a discrete search area having an associated target presence probability. Each unit may store some useful information including the probability of the presence of the object, the uncertainty level of the environment, and the situation. And combining all the units to obtain a cognitive information graph for searching.
FtargetIndicating whether the unmanned plane is in a locked state, when FtargetWhen the value is 0, the unmanned aerial vehicle is not allocated with a locking target, and the unmanned aerial vehicle flies according to the proposed algorithm; when F is presenttargetWhen the unmanned aerial vehicle is 1, the unmanned aerial vehicle is assigned to a locking target, the unmanned aerial vehicle starts to move to the target, and meanwhile, the flying height is reduced to improve the exploration probability of the load and ensure the recognition rate of target searching. PlockedAnd PtargetRespectively representing a locking probability threshold and an object probability threshold when the unit object existence probability is higher than PlockedThen, the unit is listed in a target locking sequence to allocate the corresponding unmanned aerial vehicle to perform more accurate search; when the unit object existence probability is higher than PtargetThe unit may be deemed to be present at the target point. Unmanned aerial vehicle locking state distribution principle: 1. when there is a lock unit in the target lock sequence,allocating the unmanned aerial vehicle according to the latest principle and enabling the unmanned aerial vehicle to enter a locking state; 2. when the locking unit is determined to exist at the target point or the probability of the existence of the target is lower than PlockedThe target lock sequence is removed and the drone is assigned to an unlocked state.
HthresholdSwitching threshold representing the sensing and discovery phases when flight altitude is above HthresholdWhen in the sensing stage; when the flying height is lower than HthresholdIs in the discovery phase. Because the unmanned aerial vehicle is mostly in the higher condition of flight in perception phase, the P of load this momentdAnd PfAnd if the probability map is updated according to Bayesian consistency estimation, misjudgment occurs at a high probability. Therefore, in the perception stage, a perception smoothing method is proposed for probability map updating:
where ρ represents a perceptual coefficient, and ρ may be generally 1 to Pd. And the discovery phase is updated according to the Bayesian consistency estimation.
It can be seen that the locking phase is independent of the sensing phase and the discovery phase, and is only related to its assigned locking state, while the switching between the sensing phase and the discovery phase is only related to the flying height of the drone.
Since the probability map can only reflect the probability of the existence of the target in the cell, the optimization target is that the existence probability value of the target cell is increased and the existence probability value of the target cell is not decreased, and the essence is that the certainty of the information in the cell is increased, so that the reduction of the information entropy is taken as the optimization target, namely, the uncertainty of the information in the cell is reduced.
S(t)＝-P(t)log2P(t)-(1-P(t))log2(1-P(t))
Step five: multi-unmanned aerial vehicle collaborative search target probability map fusion
Based on the search characteristics of the perceptual-lock-discovery mechanism, we propose a large-value probability map fusion method:
δgnindicating that the g unit is detected by the nth drone in the current iteration, wherein δ equals 1 to indicate detected, and δ equals 0 to indicate not detected.
Fig. 4 and 5 are simulation results of the influence of the locking phase and the sensing-discovering phase on the average target existence probability in the multi-drone collaborative search, respectively. In fig. 4, the existence probability of the target position can be greatly improved by the locking state, so that the target position can be better searched, and the capability of finding the target by the cooperative search of multiple unmanned aerial vehicles is improved. The sense-and-discover mechanism in fig. 5 improves the search performance of multiple drones better. Fig. 6 is a simulation result of the influence of the higher-value probability map fusion method on the average target existence probability in the collaborative search of multiple drones. The higher the average object existence probability, the more objects that have been searched or are about to be searched.
Fig. 7 shows that the search is performed in environments with sizes of 100, 200, 300, 400 and 500, respectively, and it can be seen that the method of the cooperative game can search for the target more effectively and has stronger robustness as the size of the environment increases.
Claims (6)
1. A multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery is characterized by comprising the following steps:
step 1: modeling the multi-unmanned aerial vehicle collaborative search environment and the flight state;
step 2: unmanned aerial vehicle searching task load modeling;
and step 3: establishing a single unmanned aerial vehicle target probability map updating mechanism;
and 4, step 4: fusing a multi-unmanned aerial vehicle collaborative search target probability graph;
and 5: and (4) solving Nash equilibrium through the potential energy game, wherein the Nash equilibrium point corresponds to a global or local optimal solution of the multi-unmanned aerial vehicle path decision in the current state.
2. The multi-unmanned aerial vehicle game cooperative search method based on perception-locking-discovery as claimed in claim 1, wherein the multi-unmanned aerial vehicle cooperative search environment and flight state modeling of step 1 specifically includes the following steps:
searching unknown environments
Equally divide the environment into Lx×LyThe units are the same in size, and each unit g is marked by (x, y);
mission drone tagging
The nth unmanned plane position (x)n，yn，hn) Identification of, wherein
hn∈[hmin，hmax]；hmin，hmaxRespectively the minimum and maximum flight heights of the unmanned aerial vehicle;
object marking requiring search
(x) for the t-th target positiont，yt) Identifying that the tth target is positioned on the ground with the height of 0 and represents that the target exists in the unit as long as the target position falls in the unit; the objective presence of an object is denoted by ω, 1 represents the object in the cell, 0 represents the object not in the cell; unmanned aerial vehicle detection results are expressed by xi, 1 represents that an object is found in a unit, and 0 represents that no object is found in the unit;
threat zone tagging
(x) for mth threat zonem，ym，Rm) Marking, the threat range is hemispherical, the sphere center of the threat range is positioned on the ground with the height of 0, and the threat radius is Rm；
All the positions of the points satisfy xn，xt，xm∈{1，2，…，LxAnd yn，yt，ym∈{1，2，…，Ly}；
{θn(t)，ηx(t) represents the flight state of the nth unmanned aerial vehicle at the time t, respectively represents the change of a course corner and a course height, is constrained by the flight performance of the unmanned aerial vehicle, and turns 45 degrees left, straight or turns 45 degrees right on the basis of the course of the unmanned aerial vehicle at the time t +1, and the maneuverability constraint conditions required to be met are as follows:
|ηn(t+1)-ηn(t)|≤ηmax
wherein eta ismaxRepresenting the maximum change value of the flight height of the unmanned aerial vehicle;
the flight conditions of the drone also satisfy the following constraints:
dij(t)≥dsafe(i，j＝1，2，…，N；i≠j)
dim(t)≥Rm(i＝1，2，…，N；m＝1，2，…，M)
wherein d isijThe distance between the ith unmanned aerial vehicle and the jth unmanned aerial vehicle is represented; dsafeRepresents a minimum safe distance between drones; dimRepresenting the distance between the ith drone and the mth threat.
3. The multi-unmanned aerial vehicle game cooperative search method based on perception-locking-discovery as claimed in claim 1, wherein the unmanned aerial vehicle search task load modeling of the step 2 specifically includes the following steps:
the photoelectric load detection model describes the detection and discovery relation of the unmanned aerial vehicle to the search target;
general probability of detection PdIs represented by PdP (ξ ═ 1| ω ═ 1); false alarm probability PfIs represented by PfP (ξ ═ 1| ω ═ 0); the photoelectric load is vertically and downwards fixedly installed, and the imaging size delta of the target on the imaging surface of the airborne photoelectric load is known from the optical imaging principle as follows:
in the formula, HsFor unmanned aerial vehicle detection height, f is photoelectric load focal length, DcIs the characteristic size of the target; height of detection HsAnd a detection radius RsThe relationship between them is:
in the formula, betasThe photoelectric load field angle; calculating the number of line pairs N of the critical dimension of the target covered on the photoelectric load imaging target surface as follows:
where b is the imaging size of the target surface of the photoelectric load, NsThe number of scanning lines is the number of photoelectric load; for the identification of a specific target, an empirical rule of a required target resolution is established, empirical data required for finding, orienting, identifying and confirming the specific target in an image are given by using a Johnson criterion, and a calculation formula of a target transfer probability function can be reversely deduced according to a data table:
4. the multi-unmanned aerial vehicle game cooperative search method based on perception-locking-discovery as claimed in claim 1, wherein the single-unmanned aerial vehicle target probability map update mechanism of step 3 specifically includes the following steps:
the cooperative search of the multiple unmanned aerial vehicles is to execute tasks through perception of the unmanned aerial vehicles on unknown environments, information interaction between the unmanned aerial vehicles and cooperative decision; during the whole task, the unmanned aerial vehicle makes a decision according to the task load and the information of the adjacent unmanned aerial vehicle, so that the search task is cooperatively executed to realize the optimal configuration; the collaborative search includes the following three parts: task load observation, information fusion and cooperative motion; before searching, each unmanned aerial vehicle associates the pre-known environmental information with a probability map, and then the unmanned aerial vehicle moves to a position with high environmental uncertainty according to an algorithm to ensure the probability of target searching; the uncertainty of a corresponding area is reduced through task load observation, and in order to further improve the searching efficiency, the unmanned aerial vehicle carries out information fusion through communication with a neighbor, so that the unmanned aerial vehicle is guided to follow-up cooperative motion; the whole process then loops until the probability distribution over all targets or the whole task space is searched for a threshold.
5. The multi-unmanned aerial vehicle game cooperative search method based on perception-locking-discovery as claimed in claim 1, wherein the multi-unmanned aerial vehicle cooperative search target probability map fusion of step 4 specifically includes the following steps:
the unmanned aerial vehicle updates the target probability map of the unmanned aerial vehicle through information interaction, and combines the information of the unmanned aerial vehicle and the acquired information to fuse the target probability map; the target probability map is represented as grid-based probability cells, wherein each cell corresponds to a discrete search area having an associated target presence probability; each unit stores some useful information including the probability of the existence of the target, the uncertainty of the environment and the situation; combining all the units to obtain a cognitive information graph for searching;
Ftargetto indicate nobodyWhether the machine is in a locked state, when FtargetWhen the value is 0, the unmanned aerial vehicle is not allocated with a locking target, and the unmanned aerial vehicle flies according to the proposed algorithm; when F is presenttargetWhen the target is 1, the unmanned aerial vehicle is assigned to lock the target, the unmanned aerial vehicle starts to move to the target, and meanwhile, the flying height is reduced to improve the exploration probability of the load and ensure the identification rate of target search; plockedAnd PtargetRespectively representing a locking probability threshold and an object probability threshold when the unit object existence probability is higher than PlockedThen, the unit is listed in a target locking sequence to allocate the corresponding unmanned aerial vehicle to perform more accurate search; when the unit object existence probability is higher than PtargetThen, the unit is determined to exist at the target point;
Hthresholdswitching threshold representing the sensing and discovery phases when flight altitude is above HthresholdWhen in the sensing stage; when the flying height is lower than HthresholdWhen in the discovery phase; because the unmanned aerial vehicle is mostly in the higher condition of flight in perception phase, the P of load this momentdAnd PfAnd (3) carrying out probability map updating by a perception smoothing method after large change occurs:
where ρ represents a perceptual coefficient, and is generally taken to be 1-Pd；
The discovery phase is updated by bayesian consistency estimation,
reducing the information entropy as an optimization target, namely reducing the uncertainty of the information in the unit;
S(t)＝-P(t)log2P(t)-(1-P(t))log2(1-P(t))。
6. the multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery as claimed in claim 1, wherein the potential energy game of the step 5 solves nash equilibrium, nash equilibrium point corresponds to global or local optimal solution of multi-unmanned aerial vehicle path decision of current state, specifically comprising the steps of:
marking a multi-unmanned plane game model as
Wherein
The method comprises the steps that a set of game participants, namely a set of task unmanned aerial vehicles is obtained;
the action set of the nth unmanned aerial vehicle is obtained; u. ofnIs the utility function of the nth unmanned plane;
if action policy set
Wherein
And
such that the utility function satisfies:
then a*Belonging to the game model
Pure strategy Nash equilibrium point of (1), wherein a-nAn action policy representing all participants except n;
if there is an accurate potential energy function phi such that
Satisfies the following conditions:
u(an′，a-n)-u(an，a-n)＝φ(an′，a-n)-φ(an，a-n)
the game is referred to as an accurate potential energy game.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111158497.6A CN114237275A (en) | 2021-09-30 | 2021-09-30 | Multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111158497.6A CN114237275A (en) | 2021-09-30 | 2021-09-30 | Multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114237275A true CN114237275A (en) | 2022-03-25 |
Family
ID=80743054
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111158497.6A Pending CN114237275A (en) | 2021-09-30 | 2021-09-30 | Multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114237275A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114697975A (en) * | 2022-04-11 | 2022-07-01 | 东南大学 | Unmanned aerial vehicle cluster distributed deployment method for enhancing land wireless coverage |
| CN116702092A (en) * | 2023-08-08 | 2023-09-05 | 中科星图测控技术股份有限公司 | Reconnaissance satellite threat degree evaluation method for comprehensive meteorological and geographic elements |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105700555A (en) * | 2016-03-14 | 2016-06-22 | 北京航空航天大学 | Potential game-based multi-unmanned aerial vehicle cooperative search method |
| CN110213771A (en) * | 2019-04-10 | 2019-09-06 | 中国人民解放军陆军工程大学 | A kind of unmanned plane relaying dispositions method based on more sensorcraft collaboratively searchings |
| CN110569855A (en) * | 2019-08-08 | 2019-12-13 | 北京航科威视光电信息技术有限公司 | Long-time target tracking algorithm based on correlation filtering and feature point matching fusion |
-
2021
- 2021-09-30 CN CN202111158497.6A patent/CN114237275A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105700555A (en) * | 2016-03-14 | 2016-06-22 | 北京航空航天大学 | Potential game-based multi-unmanned aerial vehicle cooperative search method |
| CN110213771A (en) * | 2019-04-10 | 2019-09-06 | 中国人民解放军陆军工程大学 | A kind of unmanned plane relaying dispositions method based on more sensorcraft collaboratively searchings |
| CN110569855A (en) * | 2019-08-08 | 2019-12-13 | 北京航科威视光电信息技术有限公司 | Long-time target tracking algorithm based on correlation filtering and feature point matching fusion |
Non-Patent Citations (3)
| Title |
|---|
| 吴岸平: "不确定环境下无人机区域目标搜索及载荷参数影响", 《国防科技大学学报》, pages 35 - 42 * |
| 吴楠: "面向不确定目标的多无人机协同搜索控制方法", 《计算机应用与软件》, pages 174 - 177 * |
| 潘晨: "面向 802.11b 非对称无线网络容量优化的部分重叠信道分配方案", 《通信技术》, pages 2365 - 2374 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114697975A (en) * | 2022-04-11 | 2022-07-01 | 东南大学 | Unmanned aerial vehicle cluster distributed deployment method for enhancing land wireless coverage |
| CN114697975B (en) * | 2022-04-11 | 2024-01-05 | 东南大学 | Unmanned aerial vehicle cluster distributed deployment method for enhancing land wireless coverage |
| CN116702092A (en) * | 2023-08-08 | 2023-09-05 | 中科星图测控技术股份有限公司 | Reconnaissance satellite threat degree evaluation method for comprehensive meteorological and geographic elements |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Aggarwal et al. | Path planning techniques for unmanned aerial vehicles: A review, solutions, and challenges | |
| Khan et al. | Cooperative robots to observe moving targets | |
| Wang et al. | Reconnaissance mission conducted by UAV swarms based on distributed PSO path planning algorithms | |
| Yao et al. | Gaussian mixture model and receding horizon control for multiple UAV search in complex environment | |
| Zhang et al. | An improved constrained differential evolution algorithm for unmanned aerial vehicle global route planning | |
| Recchiuto et al. | Post‐disaster assessment with unmanned aerial vehicles: A survey on practical implementations and research approaches | |
| Lin et al. | A robust real-time embedded vision system on an unmanned rotorcraft for ground target following | |
| Baik et al. | Unmanned aircraft system path planning for visually inspecting electric transmission towers | |
| CN105700555A (en) | Potential game-based multi-unmanned aerial vehicle cooperative search method | |
| Alejo et al. | Particle swarm optimization for collision-free 4d trajectory planning in unmanned aerial vehicles | |
| CN114237275A (en) | Multi-unmanned aerial vehicle game collaborative search method based on perception-locking-discovery | |
| Ru et al. | Distributed cooperative search control method of multiple UAVs for moving target | |
| Geng et al. | UAV surveillance mission planning with gimbaled sensors | |
| Farmani et al. | Tracking multiple mobile targets using cooperative unmanned aerial vehicles | |
| Droeschel et al. | Local multi-resolution surfel grids for MAV motion estimation and 3D mapping | |
| CN110823223A (en) | Path planning method and device for unmanned aerial vehicle cluster | |
| Zhou et al. | Vision-based online localization and trajectory smoothing for fixed-wing UAV tracking a moving target | |
| Zhicai et al. | A cooperative search algorithm based on improved particle swarm optimization decision for UAV swarm | |
| Hu et al. | Vision-based multi-agent cooperative target search | |
| Sadhu et al. | Aerial-DeepSearch: Distributed multi-agent deep reinforcement learning for search missions | |
| Bialas et al. | Coverage path planning for unmanned aerial vehicles in complex 3D environments with deep reinforcement learning | |
| Jung et al. | Enabling operational autonomy for unmanned aerial vehicles with scalability | |
| Zuo et al. | Overview of obstacle avoidance algorithms for UAV environment awareness | |
| Poudel et al. | Priority-aware task assignment and path planning for efficient and load-balanced multi-UAV operation | |
| El-Sallabi et al. | UAV path planning in absence of GPS signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |