A method for planning a path of a boat group target investigation based on an evolutionary algorithm

Abstract

The application discloses a kind of based on evolutionary algorithm's boat group target verification path planning method, belong to unmanned cluster autonomous control technical field, including: based on preset assumption condition, the boat group cooperative target verification task is converted into a multi-constraint condition under the class multi-traveling salesman problem;Based on particle swarm optimization algorithm, the class multi-traveling salesman problem under the multi-constraint condition is solved, so as to plan the boat group target verification path.The present application considers the load capacity of each boat, evaluates the difficulty of target verification identification, and on this basis, each boat can effectively and quickly generate the route, support dynamic task re-planning, so that the total task time is least and adaptability is strong.

Description

Technical Field

[0001] This invention belongs to the field of autonomous control technology for unmanned swarms, and more specifically, relates to a method for target verification path planning for swarms based on evolutionary algorithms. Background Technology

[0002] The single-ship path planning algorithm under multiple constraints, upon receiving the assigned task plan, can plan different routes for unmanned systems undertaking different tasks. While ensuring cooperative obstacle avoidance, the collaboratively planned routes can guarantee the formation of the unmanned system swarm and the smooth completion of its tasks. Specifically, this corresponds to minimizing the total fuel consumption of formation navigation at a given speed.

[0003] The basic concept of Particle Swarm Optimization (PSO) originates from the study of bird flock foraging behavior. It seeks the optimal solution through cooperation and information sharing among individuals in the group, and has advantages such as simple operation, easy algorithm implementation, and good robustness. The PSO algorithm can quickly converge to the optimal value at the beginning, but convergence slows down near the optimal value, and it is prone to getting trapped in local optima.

[0004] The T-ACO algorithm, based on the TSP model, addresses the Traveling Salesman Problem (TSP), a classic problem in mathematics proposed by Dr. Ramser B in 1959 based on the problem of choosing a route for a vehicle. The essence of the TSP is that a supplier delivers goods to different customers with varying needs, and the supplier must choose the shortest route without repeating any deliveries. An improved ant colony algorithm based on the T-distribution addresses the shortcomings of the basic ant colony algorithm, such as getting trapped in local optima and low optimization accuracy. During the optimization process, the T-distribution is introduced into the pheromone update principle, which helps to overcome the deficiencies of the basic ant colony algorithm. The addition of pheromone mutation to the basic ant colony algorithm increases the diversity of the ant colony, thus helping it escape the limitations of local optima. Simultaneously, the T-ACO algorithm also improves the search accuracy and convergence speed of the TSP.

[0005] Currently, utilizing unmanned surface vessels (USVs) swarms working independently can complete mine detection and clearance tasks faster than a large single platform operating alone, allowing our ships to pass through minefields more quickly. In the process of clearing mines, USVs not only need the ability to identify and destroy mines, but also a strategy to quickly traverse all mines. That is, the USV swarm must clear all pre-identified mines in a defined area of ​​water along the shortest path or in the shortest time. This is a multiple traveling salesman problem (MTSP). Existing methods for solving the MTSP are only applicable when the optimization criterion is minimizing the total sum of paths taken by all members, without considering the differences in the capabilities and target detection complexity of each platform within the swarm. Furthermore, these methods suffer from low efficiency and weak adaptability. Summary of the Invention

[0006] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention proposes a target verification path planning method for a fleet based on an evolutionary algorithm. This method considers the load capacity of each vessel and assesses the difficulty of target verification and identification. Based on this, it can effectively and quickly generate the course of each vessel, support dynamic task replanning, minimize the total task time, and enhance adaptability.

[0007] To achieve the above objectives, this invention provides a method for swarm target verification path planning based on an evolutionary algorithm, comprising:

[0008] Based on the pre-set assumptions, the task of coordinated target verification of the submarine group is transformed into a multi-traveling salesman problem under multiple constraints.

[0009] The particle swarm optimization algorithm is used to solve the multi-traveling salesman problem under multiple constraints, thereby planning the target verification path for a fleet of boats.

[0010] In some optional implementations, the preset assumptions are:

[0011] Assuming the mission area is open water and the information of each suspected target has been identified, the verification capability of each unmanned surface vessel (USV) is quantitatively calibrated. The quantified verification capability is recorded as the verification baseline. The larger the verification baseline value, the greater the corresponding verification capability. The effective verification conditions for each suspected target are defined by two indicators: cumulative revisit times: each USV reaching the preset range of the suspected target is considered a revisit; cumulative verification baseline: the sum of the verification baselines of all USVs that have reached the suspected target.

[0012] In some alternative implementations, the task of coordinating target reconnaissance of a fleet is transformed into a multi-traveling salesman problem under multiple constraints, including:

[0013] It is known that m suspected targets M1, M2, ..., M are distributed in a bounded water area E. m The set of suspected targets is M = {M1, M2, ..., M}. m The distance between any two suspected targets in} is d()∈Z+, where Z+ represents the set of positive integers;

[0014] For n unmanned surface vessels (USVs) at initial positions, n < m, where each USV is S i The verification base is Capability(i), i≤n, and the verification speed is constant; each unmanned surface vessel S i A unique route Pi = (M) Step(Si，1) M Step(Si，2) M Step(Si，j) M Step(Si，j(Si)) Step(S) i,j) represents the Sth i The suspected target number that the unmanned surface vessel (USV) is heading toward in the j-th step of its flight path, j(S i ) characterizes the Sth i The number of suspected targets that the planned route of the unmanned surface vessel (USV) must traverse; the cumulative revisit count for each suspected target is A(k), and the required value is A. R (k), k≤m; the cumulative verification base for each suspected target is AC(k), and the required value is AC(k). R (k), k≤m;

[0015] Based on the distance between suspected targets and the S of each unmanned surface vessel i The route, the cumulative number of revisits for each suspected target, and the cumulative verification base for each suspected target transform a multi-traveling salesman problem under multiple constraints into an optimization problem in mathematics.

[0016] In some alternative implementations, the optimization problem is to solve the following objective function:

[0017]

[0018] St.

[0019]

[0020] in, Indicates the Sth i The initial position of the unmanned surface vessel. Indicates the Sth i The unmanned surface vessel's final position.

[0021] In some alternative implementations, the step of using particle swarm optimization to solve the multi-traveling salesman problem under multiple constraints, thereby planning the reconnaissance path for the swarm target, includes:

[0022] The task allocation matrix is ​​obtained by describing the coordinated target verification task of the submarine group using matrix language, and the constraints are described.

[0023] The total time consumption of each scheme formed by the task allocation matrix is ​​calculated. The task allocation matrix is ​​decomposed into a weighted average of three parts: random search under multiple constraints, historical best, and inertia matrix. The random search matrix that satisfies multiple constraints is used as the mutation operator to drive the iteration.

[0024] In some optional implementations, the process of describing the coordinated target reconnaissance task of the submarine swarm using matrix language to obtain the task allocation matrix includes:

[0025] The task allocation matrix for collaborative target verification of the swarm is called Assignment. Each column of the matrix corresponds to the number of each unmanned surface vessel (USV), and each row corresponds to the number of the suspected target within the bounded waters E. The value X of Assignment[k, i] indicates whether USV numbered i in the task allocation scheme for collaborative target verification of the swarm should verify the suspected target numbered k. When X is 0, it should not be verified, and when X is 1, it should be verified.

[0026] In some alternative implementations, the description of the constraints includes:

[0027] Each row in the task assignment matrix represents the verification status of the suspected target with the corresponding number. The set of unmanned surface vessels corresponding to the column indices of all non-zero elements represents all vessel trips used to verify the suspected target. Let the set of column indices of the non-zero elements in the k-th row be K{k(1), k(2), ..., k(i), ..., k(ks)}, then the constraint condition Cond must be satisfied: length(K) represents the number of data in set K, and Capability(k(i)) represents the verification base of the kth suspected target being verified by the k(i)th unmanned surface vessel.

[0028] In some optional implementations, the calculation of the total time for each scheme formed by the task allocation matrix includes:

[0029] Each column in the task allocation matrix represents the suspected target number that the planned route of the corresponding numbered unmanned surface vessel must pass through. That is, the cluster path planning is decomposed into the path planning of each individual vessel through each column of the task allocation matrix.

[0030] The initial and final positions of each unmanned surface vessel (USV) are determined by the task assignment matrix, and the locations of the waypoints that each USV route must pass through are determined. The single-vessel route planning problem is transformed into a single-traveling salesman problem. For each column, the T-ACO algorithm T-ACO_TSP() based on the TSP model is used to solve the objective function and generate the shortest path that passes through all the necessary points.

[0031] Find the maximum value among the shortest paths corresponding to each column, and take the maximum value as the path with the longest time for all unmanned surface vessels.

[0032] In some optional implementations, the step of decomposing the task assignment matrix into a weighted average of three parts: a multi-constraint random search matrix, a historical best matrix, and an inertia matrix, with the multi-constraint random search matrix serving as the mutation operator to drive the iteration, includes:

[0033] For a random search matrix under multiple constraints, a mutation operator is generated by random generation. The random p values ​​in the k-th row of the task allocation matrix are set to 1 (p < n), indicating that the corresponding suspected target is verified by p unmanned surface vessels with the corresponding numbers. If p > A(k) and the sum of the corresponding Capability values ​​is greater than AC(k), the random search matrix is ​​considered to satisfy the constraint condition Cond. Otherwise, random numbers are regenerated and the judgment is made again until the condition is met, thus generating a random sequence that satisfies the constraint condition Cond. Then, the T-ACO algorithm based on the TSP model is used to traverse the path points to be taken by each unmanned surface vessel to obtain a local optimal solution. The length of the longest path in the entire sequence is calculated as the cost of the local optimal solution. The quality of the solution is judged by the cost value. The smaller the cost, the better the solution.

[0034] The task allocation matrix corresponding to the current local optimum is denoted as the historical optimum matrix, where the inertia matrix is ​​the result of the previous iteration. The initial values ​​of both the inertia matrix and the historical optimum matrix are set to zero. In the iteration, the values ​​of the random search matrix coefficient C1, the inertia matrix coefficient C2, and the historical optimum matrix C3 are adjusted to achieve evolutionary optimization. The solution corresponding to the task allocation matrix generated in the final iteration is defined as the final solution.

[0035] In some optional implementations, for the inertia matrix part, a reward mechanism is designed. If the cost value corresponding to the current inertia matrix is ​​better than the historical best matrix, the corresponding coefficient C2 is increased; otherwise, it is decreased. In each iteration, the value of C3 is determined according to the current iteration number, the value of C2 is calculated according to the reward mechanism, and C1 is obtained by subtracting the values ​​of C2 and C3 from 1.

[0036] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0037] This cluster path planning method takes into account the payload capacity of each vessel and assesses the difficulty of target verification and identification. Based on this, it can effectively and quickly generate the course of each vessel, support dynamic task replanning, minimize the total task time, and make it highly adaptable. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of a collaborative mine verification task modeling method for a fleet of vessels provided in an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of obtaining a task allocation matrix according to an embodiment of the present invention;

[0040] Figure 3 This is a regional search and verification task initiation phase provided by an embodiment of the present invention (from the perspective of the red team);

[0041] Figure 4 This is a regional search and verification task initiation phase (from the white side's perspective) provided in an embodiment of the present invention;

[0042] Figure 5 This invention provides an embodiment of a regional search and verification task that identifies a suspected mine target (from the red team's perspective).

[0043] Figure 6 This invention provides an embodiment of a regional search and verification task that identifies a suspected mine target (from the white side's perspective).

[0044] Figure 7 This invention provides a route planning method for a regional search and verification task (from the red team's perspective, showing the planned route of the selected unmanned surface vessel).

[0045] Figure 8 This is an embodiment of the present invention providing a route planning method for a regional search and verification task (white side view, showing the planned route of the selected unmanned surface vessel);

[0046] Figure 9 This is a schematic diagram of a process design provided by an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0048] The method for swarm target verification and path planning based on evolutionary algorithms provided by this invention, such as... Figure 1 First, a model needs to be built for the mine detection mission scenario under complete information, assuming the mission area is open water and the information of each suspected mine target has been identified. The detection capability of the unmanned surface vessel (USV) is then quantified and calibrated, denoted as the detection baseline, taking an integer value from 1 to 5; the larger the value, the greater the capability. Specifically, the effective detection conditions for each suspected mine target are defined by two indicators:

[0049] (1) Cumulative number of revisits: Each time an unmanned vessel reaches a preset range from a suspected mine target, it is considered a revisit;

[0050] (2) Cumulative verification base: The sum of the verification bases of all unmanned vessels that have reached the suspected mine target, that is, the unmanned vessels responsible for verification must have sufficient capabilities.

[0051] Based on the above assumptions, the task of coordinated mine detection by a fleet of vessels is transformed into a multi-traveling salesman problem under multiple constraints: it is known that there are m suspected mine targets M1, M2, ..., M in a bounded water area E. m The set of mines is M = {M1, M2, ..., M}. m Any M in} i M j The distance between them is d(M) i M j )∈Z+, where Z+ represents the set of positive integers;

[0052] There are n unmanned surface vessels (n < m) at the initial position, where:

[0053] (1) The verification base for each unmanned surface vessel is Capability(i), i≤n, and the verification speed is constant;

[0054] (2) S per unmanned surface vessel i A unique route Pi = (M) Step(Si，1) M Step(Si，2) M Step(Si，j) M Step(Si，j(Si)) Step(S) i ,j) characterizes S i The suspected target number that the flight path is heading towards at step j, j(S) i ) characterizes the Sth i The planned route of the unmanned surface vessel must traverse a number of suspected targets;

[0055] (3) The cumulative number of revisits for each suspected target is A(k), and the required value is A. R (k), k≤m;

[0056] (4) The cumulative verification base for each suspected target is AC(k), and the required value is AC. R (k), k≤m;

[0057] The multi-trading salesman problem under multiple constraints is then transformed into an optimization problem in mathematics, where the optimization problem is to solve the following objective function:

[0058]

[0059] St.

[0060]

[0061] in, Indicates the Sth i The initial position of the unmanned surface vessel. Indicates the Sth i The unmanned surface vessel's final position.

[0062] Then, the above optimization problem is solved based on the particle swarm optimization algorithm, and cluster path planning is performed. This can be achieved in the following way:

[0063] 1) Describe the task allocation scheme using matrix language:

[0064] Matrix language is used to describe the assignment of verification tasks. The matrix shown is the task assignment matrix for the unmanned surface vessel (USV). Each column in the matrix corresponds to the number of each USV, and each row corresponds to the number of the target within the region. The value X of Assignment[k, i] indicates whether the USV with the number i in the task assignment scheme checks the target with the number k. X is 0 when no check is performed, and X is 1 when the target is checked.

[0065] 2) Description of constraints:

[0066] Each row in the matrix represents the verification status of the target with the corresponding number. The set of unmanned surface vessels corresponding to the column indices of all non-zero elements represents all vessel trips for verifying that target. Let the set of column indices of the non-zero elements in the k-th row be K{k(1), k(2), ..., k(i), ..., k(ks)}, then the constraint Cond must be satisfied: Where length(K) represents the number of data in set K, and Capability(k(i)) represents the verification base of the kth suspected target being verified by the k(i)th unmanned surface vessel.

[0067] 3) Calculation of the total time required for the proposed solution:

[0068] Each column in the matrix represents the target number that the planned route of the corresponding unmanned surface vessel (USV) must pass through. In other words, the cluster path planning is decomposed into individual USV path planning through the columns of the matrix. For each USV, the initial and ending positions are determined, and the locations of the waypoints that the route must pass through are also determined, but their order can be adjusted. This is equivalent to transforming the individual USV path planning problem into a single traveling salesman problem (SST). The T-ACO algorithm (T-ACO_TSP()) based on the TSP model, a mature algorithm, is used to solve the objective function and generate the shortest path passing through all the necessary points. This operation is performed on each column (i.e., all USVs), and the maximum value is found. This maximum value represents the path with the longest time across all USVs in the proposed solution, i.e., the total time of the proposed solution.

[0069] 4) Evolutionary iterative design of the task assignment matrix:

[0070] The task assignment matrix is ​​decomposed into a weighted average of three parts: random search (under multiple constraints), historical best, and inertia matrix. The random search matrix that satisfies multiple constraints is used as the mutation operator to drive the iteration.

[0071]

[0072] For a random search matrix (under multiple constraints), a mutation operator is generated randomly. The p random values ​​in the k-th row of the matrix are set to 1 (p < n), indicating that the suspected target is verified by the p boats with the corresponding numbers. If p > A(k) and the sum of the corresponding Capability values ​​is greater than AC(k), then the random search matrix is ​​considered to satisfy the constraint condition Cond; otherwise, random numbers are regenerated, and the judgment is repeated until the condition is met.

[0073] like Figure 2 As shown, a random sequence satisfying the constraint Cond is generated in this way. Then, by traversing the path points to be taken by each boat using the T-ACO algorithm based on the TSP model, a locally optimal solution is obtained. The length of the longest path in the entire sequence is calculated as the cost of the solution. The solution's quality is determined by its cost; the lower the cost, the better the solution. The task allocation matrix corresponding to the current optimal solution is denoted as the history matrix.

[0074] The inertia matrix is ​​the result of the previous iteration, and the initial values ​​of both the inertia matrix and the historical best matrix are set to zero.

[0075] Evolutionary optimization is achieved by adjusting the values ​​of coefficients C1, C2, and C3 during iteration. In the initial stage, the random search component accounts for a large proportion. As the number of iterations increases, the proportion of the historical best matrix gradually increases, while the proportion of the random search component gradually decreases.

[0076] For the inertia matrix, a reward mechanism is designed: if the cost value corresponding to the current inertia matrix is ​​better than the historical best matrix, the corresponding coefficient C2 is increased; otherwise, it is decreased.

[0077] Each iteration determines the C3 value based on the current iteration (e.g., 0.2 when the iteration is less than 1000; 0.4 when the iteration is between 1000 and 2000; and 0.6 when the iteration is greater than 2000). The C2 value is calculated based on the reward mechanism. C1 is obtained by subtracting the C2 and C3 values ​​from 1.

[0078] When the number of objectives is no greater than 100, the iteration value tends to converge around 3000 iterations. The scheme corresponding to the task assignment matrix generated in the final iteration can be defined as the final scheme.

[0079] The method of the present invention will be described in detail below through a specific case.

[0080] like Figure 3 The image shows the start phase of the regional search and verification task (from the red team's perspective). Figure 4The image shows the initial phase of a regional search and verification task (from the white team's perspective). Figure 5 The image shows a suspected mine target detected during an area search and verification mission (from the red team's perspective). Figure 6 The image shows a suspected mine target detected during a regional search and verification mission (from the white side's perspective). Figure 7 The image shows the route planning for a regional search and verification mission (red team perspective, showing the selected unmanned surface vessel's planned route). Figure 8 The image shows the route planning for a regional search and verification mission (white view, showing the selected unmanned surface vessel's planned route).

[0081] Assumptions and parameter settings:

[0082] First, the unmanned surface vessel swarm forms an initial mission plan in a designated area, either independently or in small groups, and scans multiple locations of targets to be reconnoitered (the total number is known).

[0083] Second, after the target is detected during the mission, the target attributes are obtained from the simulation environment and the number of verification bases and revisited submarine types are accumulated.

[0084] Third, the cluster control system re-plans single-vessel or small-group missions based on the obtained indicators such as the number and attributes of the targets to be reconnoitered, the payload range of the unmanned surface vessel, and the number of verification bases.

[0085] Fourth, after the unmanned surface vessel swarm completes the verification of all targets in the area using the optimal collaborative scheme (completing the effective search of all targets or accumulating at least 80% of the verification base and revisit count for each target), it returns to its initial position.

[0086] Designed and parameterized according to the above assumptions, Figure 9 The process design is shown below:

[0087] The first step is to receive the combat mission information sent by the tactical decision software, where the mission attribute (pDecision->m_nTaskProperty) field is 0x95. Then, the cluster tactical decision scheme instructions sent by the tactical decision software are received, the waypoint information is parsed, and then the flight path combat instructions are sent to the cluster control software.

[0088] The second step involves the cluster control software controlling the search of friendly vessels, while the mission planning software receives target identity status information and target motion parameter information from the situation map in real time, and stores the latitude and longitude, verification base threshold, and revisit number threshold of different targets into an array.

[0089] The third step is that after all the boats (n) have completed their paths, the mission planning software receives a path tracking completion receipt instruction (n) from the cluster control software, triggering the verification phase and performing verification path planning.

[0090] The fourth step involves path planning for verification, which is then performed in the `sceneThreeFour_Other` method. The verification base number for each boat is obtained, and then they are sorted in descending order of verification base number. Three matrices are initialized: `random` (the currently generated random matrix), `history` (the best historically generated matrix), and `present` (the previously generated matrix).

[0091] Table 1

[0092]

[0093] In Table 1 above, by generating random numbers, the n values ​​in each row of the matrix are set to 1, which means that the target is verified by the n boats, and n is the number of times the target is revisited. Then, a judgment is made. If the sum of the verification base numbers of the n boats meets the verification base number threshold of the target, the random matrix of the target is established. If it does not meet the threshold, random numbers are regenerated and the judgment is made again.

[0094] This generates a random sequence that satisfies the number of revisits and the cardinality of verification. Then, by traversing the path points that each boat needs to take using the Traveling Salesman Algorithm, a locally optimal solution is obtained. The longest path length in the entire sequence is calculated and used as the cost of this solution, which is temporarily denoted as history (the current historical optimal solution).

[0095] The initial value of the present matrix is ​​all 0. A new present matrix is ​​calculated using the present, history, and random matrices. C1, C2, and C3 represent different weights to drive the subsequent transformations in the expected direction.

[0096] At the same time, when calculating the new present matrix, it may not meet the requirements for the number of revisits or the cardinality of verification. In such cases, it is necessary to re-evaluate and manually supplement the matrix with random values.

[0097] Finally, in the entire loop of 3000 iterations, the values ​​of C1, C2, and C3 are recalculated for each iteration.

[0098] After 3000 traversals, the resulting history matrix is ​​the final allocation sequence. Then, through n traveling salesmen's orders, the optimal path for each boat is obtained, and the route operation command is sent to the cluster control software.

[0099] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0100] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for target reconnaissance and path planning in a swarm based on an evolutionary algorithm, characterized in that, include: Based on the pre-set assumptions, the task of coordinated target verification of the submarine group is transformed into a multi-traveling salesman problem under multiple constraints. The particle swarm optimization algorithm is used to solve the multi-traveling salesman problem under multiple constraints, thereby planning the target verification path of the swarm. The preset assumption is as follows: Assuming the mission area is open water and the information of each suspected target has been identified, the verification capability of each unmanned surface vessel (USV) is quantified and calibrated. The quantified verification capability is recorded as the verification baseline. The larger the verification baseline value, the greater the corresponding verification capability. The effective verification conditions for each suspected target are defined by two indicators: cumulative number of revisits: each USV reaching the preset range of the suspected target is considered a revisit; cumulative verification baseline: the sum of the verification baselines of all USVs that have reached the suspected target. The transformation of the coordinated target verification task of the submarine group into a multi-traveling salesman problem under multiple constraints includes: It is known that m suspected targets M1, M2, ..., M are distributed in a bounded water area E. m The set of suspected targets is M = {M1, M2, ..., M}. m The distance between any two suspected targets in} is d()∈Z+, where Z+ represents the set of positive integers; For n unmanned surface vessels (USVs) at initial positions, n < m, where each USV is S i The verification base is Capability(i), i≤n, and the verification speed is constant; each unmanned surface vessel S i A unique corresponding route Pi = (M Step（Si，1） M Step（Si，2） M Step（Si，j） M Step（Si，j(Si)） Step(S) i ,j) represents the Sth i The suspected target number that the unmanned surface vessel (USV) is heading toward in the j-th step of its flight path, j(S i ) characterizes the Sth i The number of suspected targets that the planned route of the unmanned surface vessel (USV) must traverse; the cumulative number of revisits for each suspected target is A(k), and the required value is A. R (k), k≤m; the cumulative verification base for each suspected target is AC(k), and the required value is AC(k). R (k), k≤m; Based on the distance between suspected targets and the S of each unmanned surface vessel i The flight path, the cumulative number of revisits for each suspected target, and the cumulative verification base for each suspected target transform a multi-traveling salesman problem under multiple constraints into an optimization problem in mathematics. The method of solving the multi-traveling salesman problem under multiple constraints based on the particle swarm optimization algorithm, and planning the target verification path for the swarm, includes: The task allocation matrix is ​​obtained by describing the coordinated target verification task of the submarine group using matrix language, and the constraints are described. The total time consumption of each scheme formed by the task allocation matrix is ​​calculated. The task allocation matrix is ​​decomposed into a weighted average of three parts: random search under multiple constraints, historical best, and inertia matrix. The random search matrix that satisfies multiple constraints is used as the mutation operator to drive the iteration.

2. The method according to claim 1, characterized in that, The optimization problem is to solve the following objective function: St. ,1≤k≤m,k∈Z+ in, Indicates the Sth i The initial position of the unmanned surface vessel. Indicates the Sth i The unmanned surface vessel's final position.

3. The method according to claim 2, characterized in that, The task allocation matrix obtained by describing the coordinated target verification task of the submarine group using matrix language includes: The task allocation matrix for collaborative target verification of the swarm is called Assignment. Each column of the matrix corresponds to the number of each unmanned surface vessel (USV), and each row corresponds to the number of the suspected target within the bounded waters E. The value X of Assignment[k, i] indicates whether USV numbered i in the task allocation scheme for collaborative target verification of the swarm should verify the suspected target numbered k. When X is 0, it should not be verified, and when X is 1, it should be verified.

4. The method according to claim 3, characterized in that, The description of the constraints includes: Each row in the task assignment matrix represents the verification status of a suspected target with the corresponding number. The set of unmanned surface vessels corresponding to the column indices of all non-zero elements represents all vessel trips used to verify the suspected target. Let the set of column indices of the non-zero elements in the k-th row be denoted as . K {k(1), k(2), ..., If the expression is , …,k(ks)}, then the constraint condition Cond must be satisfied: , Represents a set K Number of data points This indicates that the k-th suspected target was detected by the 1st... The number of unmanned surface vessels used for verification.

5. The method according to claim 4, characterized in that, The calculation of the total time for each scheme formed by the task allocation matrix includes: Each column in the task allocation matrix represents the suspected target number that the planned route of the corresponding numbered unmanned surface vessel must pass through. That is, the cluster path planning is decomposed into the path planning of each individual vessel through each column of the task allocation matrix. The initial and final positions of each unmanned surface vessel (USV) are determined by the task assignment matrix, and the locations of the waypoints that each USV route must pass through are determined. The single-vessel route planning problem is transformed into a single-traveling salesman problem. For each column, the T-ACO algorithm T-ACO _TSP() based on the TSP model is used to solve the objective function and generate the shortest path that passes through all the necessary points. Find the maximum value among the shortest paths corresponding to each column, and take the maximum value as the path with the longest time for all unmanned surface vessels.

6. The method according to claim 5, characterized in that, The process of decomposing the task allocation matrix into a weighted average of three parts—random search under multiple constraints, historical best, and inertia matrix—and using the random search matrix satisfying multiple constraints as the mutation operator to drive iteration includes: For a random search matrix under multiple constraints, a mutation operator is generated by random generation. The random p values ​​in the k-th row of the task allocation matrix are set to 1, where p < n, indicating that the corresponding suspected target is verified by p unmanned surface vessels with the corresponding numbers. If p > A(k) and the sum of the corresponding Capability values ​​is greater than AC(k), then the random search matrix is ​​considered to satisfy the constraint condition Cond. Otherwise, random numbers are regenerated and the judgment is made again until the condition is met, thus generating a random sequence that satisfies the constraint condition Cond. Then, the T-ACO algorithm based on the TSP model is used to traverse the path points to be taken by each unmanned surface vessel to obtain a local optimal solution. The length of the longest path in the entire sequence is calculated as the cost of the local optimal solution. The quality of the solution is determined by the cost; the lower the cost, the better the solution. The task allocation matrix corresponding to the current local optimum is denoted as the historical optimum matrix, where the inertia matrix is ​​the result of the previous iteration. The initial values ​​of both the inertia matrix and the historical optimum matrix are set to zero. In the iteration, the values ​​of the random search matrix coefficient C1, the inertia matrix coefficient C2, and the historical optimum matrix C3 are adjusted to achieve evolutionary optimization. The solution corresponding to the task allocation matrix generated in the final iteration is defined as the final solution.

7. The method according to claim 6, characterized in that, For the inertia matrix, a reward mechanism is designed. If the cost value corresponding to the current inertia matrix is ​​better than the historical best matrix, the corresponding coefficient C2 is increased; otherwise, it is decreased. In each iteration, the value of C3 is determined according to the current iteration number, and the value of C2 is calculated according to the reward mechanism. C1 is obtained by subtracting the values ​​of C2 and C3 from 1.

Images