Multi-vehicle distributed persistent coverage trajectory planning method based on particle swarm algorithm

Abstract

The application discloses a kind of multi-vehicle distributed sustained coverage trajectory planning method based on particle swarm algorithm, utilize the distributed search characteristics of multiple unmanned vehicles, make multi-unmanned vehicle system fully perceive and collect information resources in target map, utilize the cooperation mechanism of particle swarm algorithm to realize multi-vehicle path planning, first, map gridding method supports dynamic adjustment grid size and quantity so that the application can better adapt to different size maps without changing the motion parameters of unmanned vehicle or the known function of grid;Second, the position of particle corresponds to the control input of unmanned vehicle, and the advantages and disadvantages of particle can be quickly determined by fitness function evaluation to guide the search direction of particle, speed up the convergence process, and particle swarm algorithm does not need training and has certain stability;Finally, within the scope of map target area, particle records the optimal position experienced by itself, and also learns from the optimal particle in the group, so that through group cooperation and dynamic adjustment, environmental interference can be effectively coped with.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of robot intelligent decision and control, and particularly relates to a multi-vehicle distributed persistent coverage trajectory planning method based on a particle swarm algorithm. BACKGROUND

[0002] With the continuous progress of robot perception technology, the persistent coverage problem has attracted widespread attention. The definition of the persistent coverage problem is to continuously cover and monitor a specified area or target within a certain time range. Persistent coverage aims to grasp the state and trend of the target area or object in real time, and enhance the situational awareness capability of complex environments. This special nature makes it widely used in both military and civilian scenarios. For example, in the civilian field, environmental monitoring, urban planning and emergency rescue; in the military field, border patrol, position reconnaissance and defense to ensure the safety of the camp. Persistent monitoring is also applicable to target search tasks, especially for dynamic targets. The characteristics of such targets, such as their location and movement speed, often have uncertainties, and the searcher is difficult to directly obtain relevant information. Therefore, the traditional search strategy usually adopts the way of covering the entire task area, rather than directly locating the target. Given that the target may reappear in the area that has been searched, persistent monitoring becomes a more effective strategy, which can achieve uninterrupted and continuous monitoring of the target area.

[0003] Since this type of coverage task may have certain risks, it is difficult to complete by using artificial means, so the patrol unmanned ground vehicle (UGV) emerges as the times require, which has autonomous operation capability and is suitable for various harsh and complex environments. However, with the expansion of the environmental area, the task performance of a single unmanned vehicle will be greatly reduced. To cope with this challenge, many researchers have proposed solutions for multi-vehicle cooperative operation to improve the efficiency of task execution. Compared with a single unmanned vehicle, a multi-vehicle system is composed of multiple unmanned vehicles with relatively low cost and simple structure, and can achieve better results in most cases. It is worth mentioning that the multi-vehicle system has stronger fault tolerance and parallel processing capability, and exhibits more outstanding performance in executing large-scale tasks such as extensive area monitoring or exploration, so it is an ideal choice for carrying out persistent monitoring of target areas.

[0004] The persistent coverage problem has been proved to be NP-hard in mathematics, which indicates that there is no direct analytical solution to the problem. For a long time, the methods to solve the persistent coverage problem are mainly divided into two categories. One is the method based on optimization theory, such as ant colony optimization (ACO), genetic algorithm (GA) and branch and price algorithm. The other is the method based on learning, such as deep reinforcement learning (DRL) algorithm and multi-agent reinforcement learning (MARL) algorithm. However, in a dynamic uncertain task environment, it is difficult to guarantee the stability and convergence of the learning-based method, so the evolutionary algorithm based on optimization is still the mainstream solution to such problems.

[0005] In evolutionary algorithms, particle swarm optimization (PSO) algorithm has gradually become the research direction of path planning. The PSO algorithm simulates the behavior of "bird foraging", and iteratively searches for the optimal solution in the solution space through individual exploration and group cooperation, so as to determine the current optimal or suboptimal single-step action, and then obtain the complete path and trajectory with the increase of time or step length. The PSO algorithm is suitable for global or local dynamic and high-dimensional complex problems due to its simple principle, few parameters and fast convergence speed.

[0006] The persistent coverage trajectory planning problem is a typical high-dimensional dynamic scene task. Not only the dynamic target is unknown, but also the information of the target region to be monitored is not completely known and is dynamically changing. Therefore, under the condition of partial observability of state information, how to realize the cooperative persistent monitoring trajectory planning of the target region by multiple vehicles is still a problem to be solved. SUMMARY

[0007] Therefore, the present application provides a multi-vehicle distributed persistent coverage trajectory planning method based on particle swarm optimization (PSO) algorithm. In the absence of prior information of the target region, the control signal generated by the single unmanned vehicle based on the estimation of its own field region is used for distributed cooperative exploration among multiple vehicles to realize the persistent coverage of the target region.

[0008] The multi-vehicle distributed persistent coverage trajectory planning method based on PSO algorithm provided by the present application comprises the following steps:

[0009] Step 1, constructing a known degree grid map G(t) of the target region, wherein the known degree function is g(a, b, t), (a, b) is a grid, and t is the current time;

[0010] Step 2, establishing a first motion model for trajectory planning of multiple unmanned vehicles, the first motion model is composed of a state variable s i (t), a control variable u i (t), a dynamics model and a state transition model;

[0011] Step 3, establish a trajectory optimization problem of multiple unmanned vehicles for continuous coverage according to the known grid map established in step 1 and the first motion model established in step 2, including an objective function V and constraint conditions;

[0012] Step 4, design a particle swarm solution algorithm according to the trajectory optimization problem established in step 3, including coding criteria for defining particle velocity and position, and constructing a fitness function;

[0013] Step 5, in actual use, set the initial state variables of multiple unmanned vehicles, and use the particle swarm solution algorithm to solve, to obtain the actual trajectory of multiple unmanned vehicles for continuous coverage based on the particle swarm planning strategy.

[0014] Further, the construction method of the known grid map of the target area in step 1 is:

[0015] Step 1.1, construct a grid map of m c ×m r for the target area, the state of the grid map is represented by a known function g, for a grid (a,b) with a horizontal coordinate a∈(1,m c ) and a vertical coordinate b∈(1,m r ), its known function g(a,b,t) is represented as:

[0016]

[0017] Where, t v (a,b) represents the time when the grid (a,b) is accessed by any vehicle within a set time, and its value is set to +∞ before being accessed; g v ≥0 represents the initial known degree of grid (a,b) at access time t v (a,b), η∈(0,1) is a known degree decay factor, and the known degree of the grid satisfies the condition g(a,b,t)∈(0,g v );

[0018] Step 1.2, construct the known degree of each grid as a two-dimensional matrix of m c ×m r , define the known map matrix G(t) as the known grid map.

[0019] Further, the state variable of the first motion model is represented as:

[0020] s i (t)=[p i (t),θ i (t)]

[0021] Where, i=1,2,…,N, N is the number of unmanned vehicles; p i(t) is the position of the unmanned vehicle i at the current time t, θ i (k) is the azimuth angle of the unmanned vehicle i in the inertial coordinate system;

[0022] The control variable of the first motion model is represented as:

[0023] u i (t) = [v i (t), w i (t)]

[0024] where v i (t) is the forward speed of the unmanned vehicle i at the current time t, and w i (t) is the angular speed of the unmanned vehicle i at the current time t;

[0025] The dynamic model of the first motion model is represented as:

[0026]

[0027] where Δs is the position increment, and Δθ is the azimuth angle increment;

[0028] The state transition model of the first motion model is represented as:

[0029]

[0030] where dt is the time interval, and Δs (t+1) is the state variable increment at the next time.

[0031] Further, the state variable increment Δs is calculated as: f() is the dynamic differential operator of the unmanned vehicle i, which is used to calculate the state variable increment at the next time according to the state variable s i (t) of the unmanned vehicle i at the current time and the control variable u i (t).

[0032] Further, the objective function V established in the step 3 is represented as:

[0033]

[0034] where maximizing V(t f ) represents maximizing the historical cumulative full-map awareness, u(t) is the joint control variable of the unmanned vehicle formation, and g(a, b, τ) represents the awareness of the grid (a, b) within dτ time.

[0035] The constraint condition is represented as:

[0036]

[0037] 0≤v i (t)≤v max ,-w max ≤w i (t)≤w max

[0038] where u i (t) is the input sequence of the control variable of the unmanned vehicle i, n is the terminal time, v max and w max are the maximum speed and maximum angular velocity constraints of the unmanned vehicle i, respectively.

[0039] Further, the encoding criterion of the particle speed and position in step 4 is represented as:

[0040]

[0041] where m∈[1,M] is the particle number, k∈(1,K) is the iteration algebra, the joint control variable u(t) of the unmanned vehicle formation at time t and the control change Δu(t) are encoded into the particle position and the particle speed

[0042] The update formula of the particle position and speed is:

[0043]

[0044] where w is the self-inertia factor, c1 and c2 are the individual and group learning factors, respectively, and r1 and r2∈[0,1] represent random attenuation coefficients.

[0045] The fitness function is constructed as the total map awareness at time t+dt, which is represented as:

[0046] Fitness=V(t+dt)

[0047] which is obtained by summing the awareness g(a,b,t+dt) of each grid at time t+dt.

[0048] Further, the step 5 includes:

[0049] Step 5.1, let the maximum value of the planning time be T, the time interval be d, the maximum iteration number of the particle swarm algorithm be K, and the number of particle populations in each generation be M.

[0050] Step 5.2, initialize the time t=0, and initialize the initial positions p1(0), p2(0), …, p N(0), obtain all initial state variables s(0) = [s1(0), s2(0), ···, sN(0)] of the unmanned vehicles; N (0) = [s1(0), s2(0), ···, sN(0)], initialize the grid map;

[0051] Step 5.3, record the state s(k) of the vehicle, calculate the known degree matrix G(t) of the grid map at the current time t and the objective function value V(t); initialize the position and speed of each particle according to the coding criterion;

[0052] Step 5.4, initialize the individual historical optimal position of the population by using the fitness value function Fitness individual historical optimal fitness value population historical optimal position population historical optimal fitness value Initialize the iteration number k = 0;

[0053] Step 5.5, for each particle m, calculate the position of the particle m in the k+1 generation by using the update formula and the speed Calculate the fitness value of the particle m by using the fitness value function Update the individual historical optimal position and the fitness value of the particle m

[0054] Step 5.6, update the population historical optimal position and the fitness value

[0055] Step 5.7, if k > K, execute step 5.8 to end the iteration of the particle swarm algorithm; otherwise, let k = k + 1 and execute step 5.5;

[0056] Step 5.8, perform reverse decoding by using the coding criterion to take the population historical optimal position x gbest after iteration as the real joint control variable u(t) = [u1(t), u(t), ···, uN(t)] of the unmanned vehicle at the current time t; N

[0057] Step 5.9, combine the control variable u i (t) with the state s i (t), and calculate the state variable s i (t+1) by using the state transition model;

[0058] Step 5.10, if t ≤ T, the unmanned vehicle continues to run, the current time t is updated, and step 5.3 is executed; otherwise, the unmanned vehicle terminates running, and the historical state variables [s(0), s(1), ···, s(T)] of each unmanned vehicle are output as the final unmanned vehicle formation sustained coverage trajectory.​​​

[0059] Further, the way of initializing the position and speed of each particle according to the coding criterion is random initialization.

[0060] Beneficial effects:

[0061] The present application utilizes the distributed search characteristics of multiple unmanned vehicles, so that the multiple unmanned vehicle system fully senses and collects information resources in the target map, and utilizes the cooperation mechanism of the particle swarm algorithm to plan the path of multiple vehicles. Firstly, since the map gridding method supports dynamic adjustment of the grid size and quantity, the present application can better adapt to maps of different sizes without changing the motion parameters of the unmanned vehicle or the form of the grid knowledge function. Secondly, since the position of the particle corresponds to the control input of the unmanned vehicle, the particle can be quickly determined to be good or bad through fitness function evaluation, thereby guiding the search direction of the particle and accelerating the convergence process. Moreover, the particle swarm algorithm does not need to be trained and has a certain stability. Finally, the present application has strong anti-interference ability. Within the target area range of the map, the particles record the optimal positions they have experienced, and also learn from the optimal particle in the group. Therefore, through group cooperation and dynamic adjustment, the present application can effectively cope with environmental interference.

[0062] The present application is based on the distributed motion exploration of the unmanned vehicle system, and cooperatively updates the known map, so that the system utilizes the information sharing mechanism of the particle swarm algorithm to fully utilize information and jointly realize the continuous coverage trajectory planning problem of the map area, which has important practical reference significance. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 A processing flow diagram of a multiple vehicle distributed continuous coverage trajectory planning method based on a particle swarm algorithm provided by the present application.

[0064] Figure 2 A schematic diagram of a grid map generated by a multiple vehicle distributed continuous coverage trajectory planning method based on a particle swarm algorithm provided by the present application in the embodiment.

[0065] Figure 3 A schematic diagram of a multiple unmanned vehicle continuously covering trajectory of a map area generated by a multiple vehicle distributed continuous coverage trajectory planning method based on a particle swarm algorithm provided by the present application in the embodiment.

[0066] Figure 4 A schematic diagram of a cumulative knowledge change curve over time obtained by a multiple vehicle distributed continuous coverage trajectory planning method based on a particle swarm algorithm provided by the present application in the embodiment.

[0067] Figure 5A historical coverage rate curve changing with time diagram obtained by a multi-vehicle distributed continuous coverage trajectory planning method based on a particle swarm algorithm provided in the embodiment of the present application is shown.

[0068] Figure 6 A grid visibility diagram at the end of a map area continuous coverage task obtained by a multi-vehicle distributed continuous coverage trajectory planning method based on a particle swarm algorithm provided in the embodiment of the present application is shown. DETAILED DESCRIPTION

[0069] The present application will be described in detail below with reference to the accompanying drawings and embodiments.

[0070] The core idea of the multi-vehicle distributed continuous coverage trajectory planning method based on a particle swarm algorithm provided in the present application is as follows: first, the target area is rasterized into a grid map, and then a visibility function of the grid map is introduced to quantify the coverage degree of the map; in terms of the motion model of the multi-unmanned vehicle, the state variable, the control variable, the dynamics equation and the state transition function of the unmanned vehicle are designed; on this basis, the multi-vehicle trajectory optimization problem for continuous coverage of the target area is constructed, and the objective function and the constraint condition are clarified; then, based on the particle swarm algorithm, the joint estimated state variable and the change variable of the multi-unmanned vehicle are regarded as the optimization parameters, and are coded into the position and speed of the particle to search in the solution space until the maximum number of iterations converges; the real optimal joint control variable at the current time is obtained; and then it is input into the state transition model to obtain the state variable at the next time, i.e. the position and direction information of the multi-vehicle system. This is repeated until the maximum time step is reached, and finally the optimal trajectory of the multi-unmanned vehicle system for continuous coverage is obtained. The distributed visibility map based on the unmanned vehicle system enables the system to quickly solve the trajectory planning problem and achieve full coverage of the target area as many times as possible.

[0071] The multi-vehicle distributed continuous coverage trajectory planning method based on a particle swarm algorithm provided in the present application has a processing flow as shown in Figure 1 , which specifically includes the following steps:

[0072] Step 1, constructing a visibility grid map of the target area.

[0073] Step 1.1, constructing an m c ×m r grid map for the target area, and the state of the grid map is represented by a visibility function g. For a grid (a, b) with a horizontal coordinate a∈(1, m c ) and a vertical coordinate b∈(1, m r ), at the current time t, the visibility function g(a, b, t) is represented by the following formula:

[0074]

[0075] where t v (a,b) represents the time when the grid (a,b) is accessed by any vehicle in the vicinity of the set time, and its value is set to +∞ before being accessed; g v ≥ 0 represents the time when the grid (a,b) is accessed, and its value is set to +∞ before being accessed; g v (a,b) represents the initial knowledge of the grid (a,b), and η∈(0,1) is the knowledge decay factor, and the knowledge of the grid satisfies the condition g(a,b,t)∈(0,g v ).

[0076] Step 1.2, the knowledge of each grid is constructed as a two-dimensional matrix of m c × m r , which is defined as the knowledge map matrix G(t).

[0077] Step 2, a first motion model for trajectory planning of multiple unmanned vehicles is constructed, and the first motion model is composed of state variables, control variables, a dynamics model, and a state transition model.

[0078] where the state variable is the state variable s i (t) of the unmanned vehicle i, and the expression is:

[0079] s i (t) = [p i (t), θ i (t)]

[0080] where i = 1, 2, …, N, and N is the number of unmanned vehicles; p i (t) represents the position of the unmanned vehicle i at the current time t, and θ i (k) represents the azimuth angle of the unmanned vehicle i in the inertial coordinate system.

[0081] The control variable is u i (t), and the expression is:

[0082] u i (t) = [v i (t), w i (t)]

[0083] where v i (t) is the forward speed of the unmanned vehicle i at the current time t, and w i (t) is the angular velocity of the unmanned vehicle i at the current time t.

[0084] A dynamics model is established for each unmanned vehicle i, which is used to calculate the state variable increment of the next time using the state variable and the control variable at the current time t, and the expression is:

[0085]

[0086] wherein, is the state variable increment for the next time instant; f() represents the differential operator of the unmanned vehicle dynamics, which is used to derive the state variable increment i (t) control variable u i (t) derive the state variable increment for the next time instant

[0087] A state transition model is established for each unmanned vehicle i, which is used to calculate the state variable for the next time instant given the current t time instant state variable and state variable increment, and the expression is:

[0088]

[0089] wherein, dt represents the time interval.

[0090] Step 3, a trajectory optimization problem for the continuous coverage of multiple unmanned vehicles is established according to the first motion model established in step 2, including an objective function and constraint conditions.

[0091] Specifically, the objective function V of the continuous coverage trajectory optimization problem established for the unmanned vehicle i is expressed as:

[0092]

[0093] wherein, V(t f ) represents maximizing the historical cumulative full map awareness, u(t) is the joint control variable of the unmanned vehicle formation, and g(a, b, τ) represents the awareness of the grid (a, b) within dτ time.

[0094] The constraint condition of the continuous coverage trajectory optimization problem established for the unmanned vehicle i is expressed as:

[0095]

[0096] 0≤v i (t)≤v max ,-w max ≤w i (t)≤w max

[0097] wherein, u i (t) represents the input sequence of the control variable, and n is the terminal time. v max and w max are the maximum speed and maximum angular velocity constraints of the unmanned vehicle i in the physical sense, respectively.

[0098] Step 4, a particle swarm solution algorithm is designed, including the coding criteria of particle velocity and position and the construction of fitness function, wherein the coding criteria of particle position and velocity are as follows:

[0099]

[0100] wherein, m e [1, M] represents the particle number, k e (1, K) represents the iteration algebra, the joint control variable u(t) of the unmanned vehicle formation at time t and the control change amount Au(t) are respectively encoded as the particle position Particle velocity

[0101] The update formula of the particle position and velocity is as follows:

[0102]

[0103] w in the velocity update formula of the particle represents the self inertia factor, c1 and c2 respectively represent the individual and group learning factors, and r1 and r2 e [0, 1] both represent the random attenuation coefficients.

[0104] The fitness function is constructed as follows:

[0105] Fitness = V(t + dt)

[0106] The fitness value function is defined as the full map awareness at time t + dt, which can be obtained by summing the grid awareness g(a, b, t + dt) at time t + dt.

[0107] Step 5, the process of acquiring the multi-unmanned vehicle continuous coverage trajectory planning in actual use is as follows:

[0108] Step 5.1, set parameters.

[0109] Let the maximum value of the planning time be T, the time interval be dt, the maximum iteration number of the PSO algorithm in each time interval be K, the number of particle populations in each generation be M, the self inertia factor in the velocity update formula be w, the individual and group learning factors be c1 and c2 respectively, and the random attenuation coefficients be r1 and r2.

[0110] Step 5.2, initialize the time t = 0, initialize the initial positions p1(0), p2(0), …, p N (0) of all unmanned vehicles, obtain the joint initial state variable s(0) = [s1(0), s2(0), …, s N (0)] of all unmanned vehicles, and initialize the grid map.

[0111] Step 5.3, record the state s(k) of the vehicle, calculate the awareness matrix G(t) of the grid map at the current time t and the objective function value V(t).

[0112] Step 5.4, according to the encoding rule, randomly initialize the position and velocity of each particle, the position denotes an estimated joint control variable at time t velocity denotes an estimated joint control variable at time t

[0113] Step 5.5, initialize the individual historical optimal position of the population by using the fitness value function Fitness. individual historical optimal fitness value population historical optimal position population historical optimal fitness value Initialize the iteration number k = 0.

[0114] Step 5.6, for each particle m, calculate the position of the particle m in the k+1 generation by using the update formula and velocity Calculate the fitness value of the particle m by using the fitness value function Update the individual historical optimal position of the particle m and fitness value

[0115] Step 5.7, update the population historical optimal position and fitness value

[0116] Step 5.8, if k > K, execute step 5.9 to end the iteration of the particle swarm algorithm; otherwise, let k = k + 1 and execute step 5.6.

[0117] Step 5.9, perform reverse decoding by using the encoding rule to obtain the position x gbest of the population historical optimal after iteration. N Consider the position x gbest as the real joint control variable u(t) = [u1(t), u(t), ···, u N (t)] of the unmanned vehicle at the current time t.

[0118] Step 5.10, combine the control variable u i (t) with the state s i (t), and return the state variable s i (t+1) of the next time by using the state transition model.

[0119] Step 5.11, if t ≤ T, continue the operation of the unmanned vehicle, update the current time t, and execute step 5.3; otherwise, the task reaches the predetermined maximum time length, the unmanned vehicle terminates the operation, and outputs the historical state variable [s(0), s(1), ···, s(T)] of each unmanned vehicle, which is the final unmanned vehicle formation continuous coverage trajectory.

[0120] Embodiment:

[0121] In this embodiment, a simulation environment is built by using Python language to verify the multi-vehicle distributed continuous coverage trajectory planning method based on the particle swarm optimization algorithm, and the specific process is as follows:

[0122] Simulation setting: the method provided in the present application is implemented by using Python language, in this embodiment, there are N=5 unmanned vehicles, the target area is divided into a 10x10 grid map, as shown in Figure 2 , the initial grid knowledge g v =1, the attenuation coefficient τ=0.02; let the maximum experimental time T=250, the time interval dt=1, the maximum iteration number K=100 of the PSO algorithm at each step, the particle population number M=20 remains unchanged, the self inertia factor w=0.5 in the speed update formula, the individual and group learning factors are respectively: c1=1.6, c2=1.8, and the random attenuation coefficients r1 and r2 are random numbers in the interval [0, 1].

[0123] Simulation results: Figure 3 indicates the final result, that is, the trajectory planning under the multi-vehicle distributed continuous coverage generated by the PSO algorithm. Figure 3 In the figure, the starting points of the five unmanned vehicles are random positions in the area, and the trajectories for completing the continuous coverage of the target area are cooperated. From Figure 3 , it can be seen that the unmanned vehicles respectively continue their motion in the grid map, realizing the continuous coverage of the entire area.

[0124] Figure 4 , Figure 5 respectively indicate the cumulative knowledge and the change of the historical coverage rate of the multi-unmanned vehicle system with N=5 in the time step T=250 during the training. The historical coverage rate refers to the proportion of the historical searched grid, that is, the grid with historical knowledge g(a, b, t)>0, to the total number of grid maps, which can be used to measure the continuous coverage effect. At the initial moment, the cumulative knowledge and the historical coverage rate obtained by the unmanned vehicle are very small, accompanied by the increase of the time step, the cumulative knowledge presents a fluctuating but rising trend as a whole, while the historical coverage rate gradually increases and almost approaches 100%, and finally tends to be stable.

[0125] Figure 6 shows the knowledge state of the grid map at the end of the task, and the knowledge of most of the grids is greater than 0, which shows good effect of continuous coverage.

[0126] To sum up, the above is only a preferred embodiment of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A multi-vehicle distributed continuous coverage trajectory planning method based on particle swarm optimization algorithm, characterized in that, Includes the following steps: Step 1: Construct a perceptibility raster map of the target area. The specific construction method is as follows: Step 1.1: Construct the target region A raster map, the state of which is determined by the awareness function. This means that for the x-axis is The vertical axis is grid Its degree function Represented as: ， in, Represents a grid The value is set to the time when it is accessed by any vehicle within a set time period, and before it is accessed, it is set to... ; Indicates the time of visit Grid The initial knowability, As the knowability attenuation factor, the knowability of the raster satisfies the condition. ; Step 1.2: Construct the perceptibility of each grid cell as follows A two-dimensional matrix is ​​defined as a knowability raster map matrix. As a perceptibility raster map; Step 2: Establish a first motion model for trajectory planning of multiple autonomous vehicles. The first motion model consists of state variables. Control variables It consists of a dynamic model and a state transition model; Step 3: Based on the known grid map established in Step 1 and the first motion model established in Step 2, establish a trajectory optimization problem for continuous coverage by multiple unmanned vehicles, including the objective function. and constraints; Step 4: Design a particle swarm optimization algorithm based on the trajectory optimization problem established in Step 3, including defining the encoding criteria for particle velocity and position, and constructing the fitness function; Step 5: In actual use, set the initial state variables of multiple unmanned vehicles, and use the particle swarm optimization algorithm to solve them, so as to obtain the actual trajectory continuously covered by multiple unmanned vehicles based on the particle swarm optimization strategy.

2. The multi-vehicle distributed continuous coverage trajectory planning method according to claim 1, characterized in that, The state variables of the first motion model are represented as follows: ， in, , The number of driverless cars; For driverless cars At the present moment Location, Unmanned vehicles in inertial coordinate system The azimuth angle; The control variables of the first motion model are represented as follows: ， in, For driverless cars At the current moment The forward speed, For driverless cars At the current moment angular velocity; The dynamic model of the first motion model is expressed as follows: ， in, It is a velocity vector. Angular velocity; The state transition model of the first motion model is expressed as follows: ， in, For time intervals, This represents the increment of the state variable.

3. The multi-vehicle distributed continuous coverage trajectory planning method according to claim 2, characterized in that, The increment of the state variable The calculation method is as follows: f() is the dynamic differential operator of the autonomous vehicle i, used to determine the state variables of the autonomous vehicle i at the current time. and control variables Calculate the increment of the state variable.

4. The multi-vehicle distributed continuous coverage trajectory planning method according to claim 1, characterized in that, The objective function established in step 3 Represented as: Among them, maximizing This represents maximizing the historical cumulative visibility across the entire map. For the joint control variables of the autonomous vehicle platoon, Indicates in Time grid The degree of knowability; The constraint condition is expressed as follows: in, Let i be the input sequence of the control variables for the autonomous vehicle. The final moment, and autonomous vehicles Maximum speed and maximum angular velocity constraints.

5. The multi-vehicle distributed continuous coverage trajectory planning method according to claim 1, characterized in that, The encoding criteria for particle velocity and position in step 4 are expressed as follows: ， in, Number the particles. To iterate algebraically, the autonomous vehicle formation is in Joint control variables at time 1 With control change Encoded as particle positions and particle velocity ; The formulas for updating particle position and velocity are: in, As its own inertial factor, , These are learning factors for individuals and groups, respectively. Both represent random decay coefficients; Construct the fitness function as follows: The full map visibility at any given time is represented as: From t to Within the time period, each time element The perceptibility of each grid The sum is obtained.

6. The multi-vehicle distributed continuous coverage trajectory planning method according to claim 5, characterized in that, Step 5 includes: Step 5.1: Let the maximum planning time be [value missing]. The time interval is The maximum number of iterations for the particle swarm optimization algorithm is The number of particles in each generation is ; Step 5.2, Initialize Time Initialize the initial positions of all driverless vehicles. Obtain the joint initial state variables of all autonomous vehicles. Initialize the raster map; Step 5.3: Record the vehicle's status. Calculate the current time. Awareness matrix of raster maps With the objective function value Initialize the position and velocity of each particle according to the coding criteria; Step 5.4: Use the fitness value function Initialize the historical best position of an individual in the population. Individual historical best fitness value Group's historical best position The best fitness value in history of the population The initial number of iterations is... ; Step 5.5, for each particle The updated formula is used to calculate the first... Particles of the generation Location With speed The fitness value function is used to calculate the particle fitness value Update particles Individual historical best position and fitness value ; Step 5.6: Update the group's historical best position and fitness value ; Step 5.7, if If so, proceed to step 5.8 to end the particle swarm algorithm iteration; otherwise, let... And proceed to step 5.5; Step 5.8: Perform reverse decoding using the encoding criteria to determine the position of the historical best of the group after iteration. Consider the current moment Real joint control variables of autonomous vehicles ; Step 5.9, combining control variables With state The state variables at the next time step are calculated using a state transition model. ; Step 5.10, if The driverless car will continue operating and update the current time. Execute step 5.3; otherwise, the autonomous vehicle will terminate operation and output the historical state variables of each autonomous vehicle. As the final unmanned vehicle platoon continuously covers the trajectory.

7. The multi-vehicle distributed continuous coverage trajectory planning method according to claim 6, characterized in that, The method for initializing the position and velocity of each particle according to the coding criteria is random initialization.

Images