A Method and System for Joint Planning of Heterogeneous UAV Swarm Missions and Flight Paths Based on Feasibility Feedback

Abstract

This invention discloses a method and system for joint planning of tasks and trajectories in heterogeneous UAV swarms based on feasibility feedback. The method includes: acquiring task set information, UAV swarm status information, and task environment information; constructing a multi-dimensional matching degree matrix between UAVs and tasks; generating an initial planning population based on the matrix, including task allocation codes and trajectory key point codes; performing feasibility closed-loop verification on individual planning entities, calculating the estimated flight energy consumption and task completion time based on the generated three-dimensional flight trajectory, and filtering based on remaining battery power constraints and task time window constraints; modulating pheromone concentration using the feasibility verification results to guide the search direction; and outputting a joint planning scheme for task allocation and three-dimensional trajectory that satisfies the flight constraints. This invention effectively solves the problem of infeasibility caused by the separation of task allocation and trajectory planning, improving the actual executability and optimization efficiency of the planning scheme.

Description

Technical Field

[0001] This invention relates to the field of intelligent control and collaborative planning technology for unmanned aerial vehicles (UAVs), and in particular to a method and system for joint planning of heterogeneous UAV swarm tasks and trajectories based on feasibility feedback. Background Technology

[0002] With the rapid development of unmanned aerial vehicle (UAV) technology, multi-UAV collaborative execution of complex tasks has been widely applied in disaster relief, regional reconnaissance, logistics transportation, and emergency support. In such applications, a swarm of multiple heterogeneous UAVs is typically required to collaboratively handle multiple target tasks. This mainly involves two core decision-making issues: first, how to rationally allocate tasks to different UAVs based on mission requirements and UAV status; and second, how to plan feasible flight paths for each UAV that meet flight constraints.

[0003] In existing technologies, multi-UAV mission planning typically employs a two-stage serial decoupling method of "task allocation—path planning." First, tasks are allocated based on static indicators such as task priority and UAV location, determining the task sequence for each UAV. Then, path planning is performed for each UAV based on the allocation results. However, this method has significant limitations: in the task allocation stage, it often only considers static factors such as task priority and straight-line distance, failing to fully account for energy consumption constraints, mission time window constraints, and the nonlinear effects of complex environmental obstacles during actual flight. This "decoupling" leads to the "optimal" allocation scheme generated in the first stage frequently failing during the second-stage path planning due to insufficient energy or unreachability, forcing the system to replan and reducing the efficiency and reliability of mission execution.

[0004] Furthermore, although existing technologies have attempted joint planning using swarm intelligence algorithms such as genetic algorithms and ant colony algorithms, these algorithms typically handle constraints only through penalty terms in the fitness function during the iterative search process, lacking real-time closed-loop feedback on physical constraints. This often leads to the algorithms wasting significant computational resources in infeasible solution spaces and easily getting trapped in local optima due to improperly set penalty coefficients, making it difficult to generate globally optimal solutions that meet actual flight conditions.

[0005] Therefore, how to break down the disconnect between task allocation and trajectory planning during the planning process, and how to combine physical constraints with real-time closed-loop feedback guidance during the search process, so as to efficiently obtain a joint planning scheme that meets actual flight conditions, is a key technical problem that urgently needs to be solved in the field of multi-UAV collaborative planning. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for joint planning of heterogeneous UAV swarm missions and trajectories based on feasibility feedback, thereby solving the aforementioned problems existing in the prior art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] Firstly, a method for joint planning of heterogeneous UAV swarm missions and flight paths based on feasibility feedback includes the following steps:

[0009] S1. Obtain task set information, drone cluster status information, and task environment information;

[0010] S2. Construct a multi-dimensional matching degree matrix between UAVs and tasks based on task set information, UAV cluster status information, and task environment information.

[0011] S3. Generate an initial planning population based on a multidimensional matching degree matrix, where each planning individual is represented by a joint coding structure, which includes task allocation coding and track key point coding.

[0012] S4. Perform a closed-loop feasibility check on the individual drones in the planning population, including: determining the task execution sequence for each drone based on the task allocation code; generating a three-dimensional flight track for the corresponding drone based on the task execution sequence, track key point code, and task environment information; calculating the expected flight energy consumption and expected task completion time for each drone based on the three-dimensional flight track; comparing the expected flight energy consumption with the remaining battery power constraint of the corresponding drone, and comparing the expected task completion time with the time window constraint of the corresponding task, to obtain the feasibility check results for the individual drones.

[0013] S5. Based on the feasibility verification results, screen and adjust the planning individuals. For planning individuals that do not meet the constraints, apply fitness penalties or delete them. For planning individuals that meet the constraints, retain them.

[0014] S6. Construct a joint optimization objective function based on the planning individuals that have passed the feasibility verification;

[0015] S7. The population is iteratively optimized by a joint search strategy that combines genetic evolution operations and pheromone update mechanisms. The pheromone concentration of the task allocation code and track key point code combination is modulated using the feasibility verification results. Based on the updated pheromone concentration, the selection, crossover and mutation operations of the planned individuals are performed to generate a new planned population.

[0016] Modulating pheromone concentration using feasibility verification results specifically includes: increasing the corresponding pheromone concentration for task allocation codes and track key point codes that pass feasibility verification; decreasing the corresponding pheromone concentration for task allocation codes and track key point codes that fail feasibility verification; wherein, pheromone concentration serves as heuristic information to participate in the selection operation of planning individuals, guiding the population to converge toward the feasible solution space.

[0017] S8. When the preset termination conditions are met, output the task allocation and three-dimensional trajectory joint planning scheme that meets the flight constraints.

[0018] In an exemplary embodiment, a local repair step is further included between steps S4 and S5:

[0019] When the expected flight energy consumption of a certain planning individual exceeds the remaining power constraint of the UAV but the excess is less than the preset threshold, a local flight path replanning is triggered for the planning individual. The local flight path is regenerated by adjusting the code of the key points of the flight path, and the feasibility closed-loop verification is re-executed based on the replanning result.

[0020] The preset threshold is dynamically determined based on the battery discharge characteristic curve of the drone and the task priority. For power consumption during the discharge plateau period or high-priority tasks, a larger threshold is used to increase the chance of replanning; for power consumption at the end of the discharge period or low-priority tasks, a smaller threshold is used to strictly screen feasible solutions.

[0021] In an exemplary embodiment, the trajectory key point encoding is used to represent the three-dimensional sub-target nodes that guide the UAV's flight. During the trajectory generation process, the path planner starts from the current position of the UAV, passes through the sub-target nodes corresponding to the trajectory key point encoding in sequence, and finally reaches the mission point. By connecting the sub-target nodes in the three-dimensional grid environment, a wingable trajectory that meets obstacle avoidance constraints is generated. The trajectory key point encoding indirectly guides the flight path shape by controlling the spatial distribution of the sub-target nodes.

[0022] In an exemplary embodiment, the multidimensional matching degree matrix comprehensively considers at least two of the following factors: the compatibility between the drone type and the mission type, the distance between the drone's current location and the mission point, the matching degree between the drone's payload capacity and the mission payload requirements, and the matching degree between the drone's remaining battery power and the estimated flight energy consumption.

[0023] In an exemplary embodiment, the joint optimization objective function is used to perform weighted joint optimization of task execution benefits, total flight energy consumption of the UAV swarm, and task completion time span, under the premise of satisfying feasibility constraints.

[0024] In an exemplary embodiment, the output task allocation and three-dimensional trajectory joint planning scheme is a Pareto optimal solution set consisting of multiple planning schemes obtained through a multi-objective optimization algorithm that are mutually independent in the three objective dimensions of task execution benefits, total energy consumption of UAV swarm flight, and task completion time span.

[0025] Secondly, a heterogeneous UAV swarm mission and trajectory joint planning system based on feasibility feedback includes:

[0026] The data acquisition module is used to acquire task set information, UAV cluster status information, and task environment information;

[0027] The matching degree construction module is used to build a multi-dimensional matching degree matrix between the drone and the mission;

[0028] The population generation module is used to generate an initial planned population using a joint coding structure.

[0029] The feasibility verification module is used to generate a three-dimensional flight track based on the task allocation code and the track key point code, and output the feasibility verification results based on the remaining power constraints and the task time window constraints.

[0030] The joint search module is used to perform fitness adjustment and pheromone concentration update based on the feasibility verification results, and to iteratively generate a new planning population through selection, crossover and mutation operations. Among them, the joint search module uses the feasibility verification results to modulate the pheromone concentration of the combination of task allocation code and track key point code, and guides the population to converge toward the feasible solution space.

[0031] The command output module is used to output a joint planning scheme for task allocation and three-dimensional trajectory that meets the constraints.

[0032] Thirdly, a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0033] The beneficial effects of this invention are:

[0034] By introducing a feasibility feedback mechanism into the joint optimization process of task allocation and trajectory planning, planning individuals that do not meet flight constraints are promptly screened during the search process, and the pheromone concentration is modulated using the verification results to guide the search direction. This makes the planning process more in line with the actual flight conditions of UAVs, solves the problem of infeasible solutions in traditional two-stage methods, and improves optimization efficiency and the quality of planning results. Attached Figure Description

[0035] Figure 1 A flowchart of the method provided in an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the system module architecture and joint coding structure provided in the embodiments of the present invention;

[0037] Figure 3 This is a schematic diagram of drone task allocation provided in an embodiment of the present invention. Detailed Implementation

[0038] 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. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0039] In one exemplary embodiment, refer to Figure 1 , Figure 2 and Figure 3 As shown, a method for joint planning of heterogeneous UAV swarm missions and trajectories based on feasibility feedback is provided, including the following steps S1 to S8. Wherein:

[0040] Firstly, a method for joint planning of heterogeneous UAV swarm missions and flight paths based on feasibility feedback includes the following steps:

[0041] S1. Obtain task set information, drone cluster status information, and task environment information.

[0042] S2. Construct a multi-dimensional matching degree matrix between UAVs and tasks based on task set information, UAV cluster status information, and task environment information.

[0043] S3. Generate an initial planning population based on a multidimensional matching degree matrix, where each planning individual is represented by a joint coding structure, which includes task allocation coding and track key point coding.

[0044] S4. Perform a closed-loop feasibility check on the individual drones in the planning population, including: determining the task execution sequence for each drone based on the task allocation code; generating a three-dimensional flight track for the corresponding drone based on the task execution sequence, track key point codes, and task environment information; calculating the expected flight energy consumption and expected task completion time for each drone based on the three-dimensional flight track; comparing the expected flight energy consumption with the remaining battery power constraint of the corresponding drone, and comparing the expected task completion time with the time window constraint of the corresponding task, to obtain the feasibility check results for the individual drones.

[0045] S5. Based on the feasibility verification results, screen and adjust the planning individuals. For planning individuals that do not meet the constraints, apply fitness penalties or delete them. For planning individuals that meet the constraints, retain them.

[0046] S6. Construct a joint optimization objective function based on the planning individuals that have passed the feasibility verification.

[0047] S7. A joint search strategy combining genetic evolution operations and pheromone update mechanisms is adopted to iteratively optimize the population. The pheromone concentration of the task allocation code and track key point code combination is modulated using the feasibility verification results. Based on the updated pheromone concentration, the selection, crossover and mutation operations of planned individuals are performed to generate a new planned population.

[0048] Modulating pheromone concentration using feasibility verification results specifically includes: increasing the corresponding pheromone concentration for task allocation codes and track key point codes that pass feasibility verification; decreasing the corresponding pheromone concentration for task allocation codes and track key point codes that fail feasibility verification; wherein, pheromone concentration serves as heuristic information to participate in the selection operation of planning individuals, guiding the population to converge toward the feasible solution space.

[0049] S8. When the preset termination conditions are met, output the task allocation and three-dimensional trajectory joint planning scheme that meets the flight constraints.

[0050] Implementing steps S1 to S8 above allows heterogeneous drone swarms to efficiently generate task allocation and trajectory planning schemes that meet actual flight feasibility under complex constraints.

[0051] To achieve the above objectives, this invention provides a method for joint planning of heterogeneous UAV swarm missions and flight paths based on feasibility feedback, comprising the following steps:

[0052] First, acquire task set information, drone swarm status information, and task environment information. The task set information includes task location, task type, task priority, and task time window constraints. The drone swarm status information includes drone type, current location, remaining battery power, flight speed, and payload capacity. The task environment information includes no-fly zones, obstacle zones, or threat zones.

[0053] A multi-dimensional matching degree matrix between the UAV and the task is constructed based on task information and UAV status information. This matching degree matrix is ​​used to quantify the degree of adaptability of the UAV in performing the task, and its calculation method is as follows:

[0054]

[0055] in:

[0056] This represents the overall matching degree of the i-th drone performing the j-th task;

[0057] Indicates the compatibility between drone type and mission type;

[0058] This indicates the distance matching item between the drone's current location and the mission point:

[0059] This indicates the matching items between the drone's payload capacity and mission requirements;

[0060] This indicates the degree of matching between the drone's remaining battery power and the estimated flight energy consumption;

[0061] Let be the weight coefficient, and satisfy...

[0062] ;

[0063] The distance matching item is calculated as follows:

[0064] ;

[0065] After obtaining the multidimensional matching degree matrix, an initial planning population is generated. Each planning individual is represented using a joint coding structure, which includes task allocation coding and trajectory key point coding. Task allocation coding represents the mapping relationship between the task and the UAV; trajectory key point coding represents the 3D sub-target nodes in the UAV's flight path. These nodes are not precise waypoints that must be crossed, but rather serve as guiding sub-targets for path search. The path planner starts from the UAV's current position in a 3D grid environment, sequentially passes through these sub-target nodes, and finally reaches the task point. By searching and connecting between grid nodes, a 3D flight trajectory is generated that satisfies obstacle avoidance requirements and passes through the sub-target guidance area. By adjusting the coordinate values ​​of the trajectory key point coding, the flight path shape can be flexibly adjusted while maintaining the overall task allocation, thereby achieving joint optimization of task allocation and trajectory planning.

[0066] Subsequently, a closed-loop feasibility check is performed on the planned individual drones. The system determines the drone mission execution sequence based on the task allocation code and generates a corresponding 3D flight path for the drone by combining the key point codes of the flight path. The estimated flight energy consumption of the drone is calculated based on the generated flight path; the calculation model can be expressed as follows:

[0067] ;

[0068] in:

[0069] Let be the estimated flight energy consumption of the i-th UAV;

[0070] This is the length of the flight path;

[0071] The number of turns or the cost of turning;

[0072] The cost of traversing threatened areas;

[0073] It is the energy consumption weighting coefficient, and satisfies .

[0074] Based on the projected flight energy consumption and mission completion time, the feasibility of the planned individuals is verified by comparing them with the constraints of the UAV's remaining battery power and the mission time window. For planned individuals that do not meet the constraints, a fitness penalty or elimination mechanism is used to reduce their probability of appearing in subsequent searches. The path cost function is used in the trajectory generation process.

[0075]

[0076] in:

[0077] Path cost

[0078] L is the path length.

[0079] R represents the cost of the threat area.

[0080] Cost of steering angle

[0081] Weighting coefficients

[0082] Based on the individuals that have passed feasibility verification, a joint optimization objective function is constructed to comprehensively optimize the task execution benefits, the total energy consumption of the UAV swarm flight, and the task completion time span. The objective function can be expressed as:

[0083] ;

[0084] in:

[0085] F represents the comprehensive evaluation value of the individual being planned;

[0086] R represents the total revenue from task execution;

[0087] E represents the total flight energy consumption of the drone swarm;

[0088] T represents the time span for task completion;

[0089] These are weighting coefficients used to adjust the importance of different optimization objectives.

[0090] During the optimization process, a joint search strategy combining genetic evolution operations and pheromone update mechanisms is adopted to generate a new generation of planned population through selection, crossover, and mutation operations.

[0091] In each iteration, the results of the feasibility closed-loop verification adjust the search direction by influencing the pheromone update rules. It is important to note that this invention does not solely rely on the positive feedback mechanism of pheromones; instead, it integrates pheromone concentration as heuristic information into the selection operations of the genetic algorithm, forming a hybrid search strategy of "pheromone guidance + genetic exploration." On one hand, differences in pheromone concentration increase the probability of retaining and propagating high-fitness coding combinations, guiding the population towards the feasible solution space. On the other hand, crossover and mutation operations in the genetic algorithm continuously introduce new coding combinations, maintaining population diversity and effectively avoiding premature convergence caused by excessive pheromone accumulation.

[0092] When establishing the feasibility-pheromone mapping function, a continuous feedback mechanism is used instead of a simple binary reward and punishment mechanism:

[0093] ;

[0094] in:

[0095] The pheromone concentration of the encoded combination k at the t-th iteration;

[0096] The pheromone evaporation coefficient;

[0097] This is to increase the amount of pheromones.

[0098] The pheromone increment is determined based on the feasibility verification results:

[0099]

[0100] In the next iteration, the probability of the code combination being selected is calculated according to the following formula:

[0101] ;

[0102] in:

[0103] Let k be the probability that the code combination k is selected.

[0104] The concentration of pheromones;

[0105] For heuristic information;

[0106] These are weight parameters;

[0107] This is the set of candidate encoding combinations.

[0108] When the preset termination condition is met, the system outputs a joint planning scheme for task allocation and 3D trajectory that satisfies the constraints of remaining battery power and task time window. The planning result can be represented as a solution set composed of multiple non-dominated solutions to provide a planning scheme that balances different objectives, thereby improving the flexibility of planning decisions.

[0109] By introducing a feasibility feedback mechanism into the joint optimization process of task allocation and trajectory planning, this invention can promptly filter out planning entities that do not meet flight constraints during the search process, making the planning process more consistent with the actual flight conditions of the UAV. Simultaneously, a joint coding structure enables collaborative optimization of task allocation and trajectory planning, and a pheromone feedback mechanism adjusts the search direction, thereby obtaining a joint task allocation and trajectory planning scheme that satisfies the flight constraints.

[0110] Secondly, a local repair step is also included between steps S4 and S5:

[0111] When the expected flight energy consumption of a certain planning individual exceeds the remaining power constraint of the UAV but the excess is less than the preset threshold, a local flight path replanning is triggered for the planning individual. The local flight path is regenerated by adjusting the code of the key points of the flight path, and the feasibility closed-loop verification is re-executed based on the replanning result.

[0112] The preset threshold is dynamically determined based on the battery discharge characteristic curve of the drone and the task priority. For power consumption during the discharge plateau period or high-priority tasks, a larger threshold is used to increase the chance of replanning; for power consumption at the end of the discharge period or low-priority tasks, a smaller threshold is used to strictly screen feasible solutions.

[0113] In an exemplary embodiment, the trajectory key point encoding is used to represent the three-dimensional sub-target nodes that guide the UAV's flight. During the trajectory generation process, the path planner starts from the current position of the UAV, passes through the sub-target nodes corresponding to the trajectory key point encoding in sequence, and finally reaches the mission point. By connecting the sub-target nodes in the three-dimensional grid environment, a wingable trajectory that meets obstacle avoidance constraints is generated. The trajectory key point encoding indirectly guides the flight path shape by controlling the spatial distribution of the sub-target nodes.

[0114] In an exemplary embodiment, the multidimensional matching degree matrix comprehensively considers at least two of the following factors: the compatibility between the drone type and the mission type, the distance between the drone's current location and the mission point, the matching degree between the drone's payload capacity and the mission payload requirements, and the matching degree between the drone's remaining battery power and the estimated flight energy consumption.

[0115] In an exemplary embodiment, the joint optimization objective function is used to perform weighted joint optimization of task execution benefits, total flight energy consumption of the UAV swarm, and task completion time span, under the premise of satisfying feasibility constraints.

[0116] In an exemplary embodiment, the output task allocation and three-dimensional trajectory joint planning scheme is a Pareto optimal solution set consisting of multiple planning schemes obtained through a multi-objective optimization algorithm that are mutually independent in the three objective dimensions of task execution benefits, total energy consumption of UAV swarm flight, and task completion time span.

[0117] Secondly, a heterogeneous UAV swarm mission and trajectory joint planning system based on feasibility feedback includes:

[0118] The data acquisition module is used to acquire task set information, UAV cluster status information, and task environment information;

[0119] The matching degree construction module is used to build a multi-dimensional matching degree matrix between the drone and the mission;

[0120] The population generation module is used to generate an initial planned population using a joint coding structure.

[0121] The feasibility verification module is used to generate a three-dimensional flight track based on the task allocation code and the track key point code, and output the feasibility verification results based on the remaining power constraints and the task time window constraints.

[0122] The joint search module is used to perform fitness adjustment and pheromone concentration update based on the feasibility verification results, and to iteratively generate a new planning population through selection, crossover and mutation operations. Among them, the joint search module uses the feasibility verification results to modulate the pheromone concentration of the combination of task allocation code and track key point code, and guides the population to converge toward the feasible solution space.

[0123] The command output module is used to output a joint planning scheme for task allocation and three-dimensional trajectory that meets the constraints.

[0124] Thirdly, a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0125] Example 1:

[0126] This embodiment uses earthquake disaster relief in a city as an application scenario to verify the technical effect of the present invention. The task area is set as a 5km × 5km urban core area, including 5 no-fly zones such as collapsed buildings and fire zones. The task set includes 7 task points, specifically: 3 survivor search tasks (high priority, generous time window), 2 life support delivery tasks (high priority, strict time window ±1min), and 2 disaster reconnaissance tasks (medium priority, generous time window). The drone swarm consists of 5 heterogeneous drones, including 3 small rotary-wing drones (mainly used for reconnaissance missions, maximum range 12km, endurance 20min) and 2 medium-sized fixed-wing drones (mainly used for supply delivery missions, maximum range 25km, endurance 40min).

[0127] Step 1: Data Collection and Matching Degree Matrix Construction

[0128] The ground station system collects task set information (task point coordinates, task type, priority, time window constraints, and material requirements) and UAV swarm status information (position, battery level, speed, and payload capacity) via communication links. The matching degree calculation module constructs a 5×7 multidimensional matching degree matrix. , representing the overall matching degree of drone i performing task j, is calculated using the following formula:

[0129]

[0130] TypeMatch represents the type matching score (1 for the same type and 0.5 for different types).

[0131] The normalized distance matching degree ( (where Euclidean distance is used), CapMatch represents the load matching degree, and EnergyMatch represents the battery matching degree. The weighting coefficients are set to 0.3, 0.3, 0.2, and 0.2 respectively in this embodiment, based on the comprehensive requirements of task adaptability in the rescue scenario. It should be noted that these weighting coefficients are only exemplary values ​​and can be dynamically adjusted according to different task preferences (such as placing more emphasis on distance factors or battery power factors) in actual applications. This demonstrates the adaptability of the method of the present invention to different application scenarios.

[0132] Step 2: Population Initialization

[0133] The population initialization module generates 80 initial individuals based on the matching degree matrix using a roulette wheel strategy. Each individual contains two parts of encoding: a task allocation encoding. Indicates the drone number performing the mission; track key point code. , representing the coordinates of key nodes that each drone must pass through in its flight path. The number of key nodes is dynamically determined based on the distance to the mission point and the complexity of the environment. In this embodiment, 3-5 key points are set for each flight path.

[0134] Step 3: Feasibility Closed-Loop Verification

[0135] For each individual, the feasibility verification module calls the trajectory generator (improved). The algorithm generates a complete 3D trajectory. The trajectory generator searches the 3D grid map for the optimal path from the UAV's current position to the mission point, sequentially passing through key points. The path cost function simultaneously considers flight distance, threat area crossing cost, and turning energy consumption.

[0136]

[0137] in:

[0138] Path cost

[0139] L is the path length.

[0140] R represents the cost of the threat area.

[0141] Cost of turning

[0142] Weighting coefficient

[0143] Based on the generated complete flight path, calculate the estimated energy consumption E and estimated mission completion time T for each drone. Then, compare the estimated energy consumption with the drone's remaining battery power threshold. Expected task completion time and task time window threshold A comparison is made. For individuals who do not meet the constraints, instead of simply eliminating them directly, a "soft feedback" mechanism is introduced:

[0144] like And the excess is more than 5%, or If the excess exceeds 10% of the time window width, it is considered a serious violation, and the individual's fitness value is set to zero (elimination).

[0145] like However, the excess is within 5%, or However, if the excess is within 10% of the time window width, a local replanning of the flight path will be triggered. It should be noted that the 5% and 10% thresholds are not fixed and can be dynamically adjusted based on the drone's battery discharge characteristics in a real system. For example, when the drone is in the battery discharge plateau phase (voltage stable stage), a short period of slight over-consumption will not cause a sudden drop in power; in this case, the threshold can be appropriately relaxed to 8%, allowing more replanning opportunities. When the drone is in the final stage of discharge, even a small over-consumption can lead to complete power depletion; in this case, the threshold should be tightened to 3%. Furthermore, for high-priority tasks, the threshold can be appropriately relaxed to ensure that tasks are executed as much as possible; for low-priority tasks, the threshold should be strictly enforced to ensure the overall safety of the cluster.

[0146] Specifically, under the premise of fixing the UAV mission allocation sequence, the degree of freedom for adjusting the track key point encoding is released, and within the three-dimensional grid neighborhood between the current position and the next target point, the following is adopted: The Lite algorithm performs local fine-tuning of keypoint coordinates to regenerate a more energy-efficient flight path. If the constraints still cannot be met after local replanning, the individual is discarded; if the constraints are met, it is retained and its path code is updated.

[0147] If both energy consumption and time window constraints are met, the individual can be directly retained for subsequent optimization.

[0148] Step 4: Construction of Joint Optimization Objective Function

[0149] Based on the individuals that pass the verification, a joint optimization objective function is constructed:

[0150]

[0151] in, To calculate the total reward for mission execution based on mission priority and completion quality, in this embodiment, the priority of the survivor search mission is set to 0.5, the priority of the material delivery mission is set to 0.3, and the priority of the disaster reconnaissance mission is set to 0.2. Total energy consumption for cluster flight; A penalty is imposed for the time span of task completion. When the time span of all tasks exceeds 5 minutes, a non-linear penalty is applied, and the penalty magnitude increases with the time exceeding the time. The weighting coefficients are set to 0.4, 0.4, and 0.2 respectively in this embodiment, based on a comprehensive consideration of benefits, energy consumption, and time span. In practical applications, these coefficients can be adjusted according to the preferences of the mission commander; for example, the weight of time span can be appropriately increased in emergency rescue scenarios.

[0152] Step 5: Iterative Optimization

[0153] An evolutionary algorithm based on pheromone mechanisms is used for iterative optimization. A new generation of the population is generated through selection, crossover, and mutation operations. The number of iterations is preset to 350 generations, and the population size remains constant at 80. In each iteration, the results of the feasibility closed-loop verification influence the pheromone update rules to adjust the search direction.

[0154] When establishing the feasibility-pheromone mapping function, a continuous feedback mechanism is used instead of a simple binary reward and punishment mechanism:

[0155]

[0156] in, and The attenuation coefficient is... and For exceeding the limit, The reward coefficient is greater than 1. This serves as a penalty measure. Through this continuous feedback mechanism, individuals carrying excellent task allocation genes but only slightly violating constraints are given a moderate penalty rather than being completely eliminated. This guides the search towards the feasible region while maintaining population diversity and avoiding missing potential excellent gene combinations.

[0157] The pheromone concentration update rules are as follows:

[0158]

[0159] in, Here is the pheromone evaporation coefficient. In the next iteration, the probability of the coded combination being selected is calculated based on the updated pheromone concentration:

[0160] ;

[0161] in, This is heuristic information (calculated based on the matching degree matrix). and These are the weight parameters.

[0162] The selection operation employs a tournament selection strategy, retaining individuals with the top 20% fitness; the crossover operation uses two-point crossover for task assignment encoding and simulated binary crossover for track key point encoding; the mutation operation randomly perturbs the encoding with a probability of 0.1.

[0163] Step 6: Output the Pareto optimal solution set

[0164] After iteration, Pareto optimal solutions are extracted from the last generation of the population based on non-dominated sorting, forming the Pareto front distribution. This embodiment outputs three sets of Pareto optimal solutions for the commander to choose from:

[0165] Option A: The most energy-efficient option, with a total energy consumption of 487 kWh and a task completion time of 28 minutes;

[0166] Option B: The fastest completion option, with a total energy consumption of 532 kWh and a task completion time of 21 minutes;

[0167] Option C: Balanced solution, total energy consumption 501 kWh, task completion time 24 min.

[0168] Step 7: Implementation of the Plan

[0169] Based on the actual situation on site, the commander selected Plan C for execution. The command issuance module converted the optimized plan into standard MAVLink commands and sent them to each UAV via data link. All UAVs flew along the planned flight path, successfully avoiding all no-fly zones, and completed all tasks on time. No UAVs returned due to insufficient power, achieving a 100% mission success rate.

[0170] To verify the technical effects of this invention, a simulation platform incorporating complex urban environmental obstacles was constructed for comparative experiments. To highlight the unique contribution of the core feature of this invention—the "feasibility feedback mechanism"—three comparative groups were set up: Control group A employed the classic two-stage "allocation followed by planning" method, where task allocation and path planning were performed independently; Control group B used a joint encoding optimization method without introducing feasibility feedback, i.e., only joint optimization without a feedback loop; the invention group fully adopted the joint optimization and feasibility feedback mechanism of this invention.

[0171] The experimental parameters were set as follows: The mission scenario was set in a 10km×10km three-dimensional urban area, including 5 no-fly zones. The drone swarm consisted of 6 heterogeneous drones, including 4 rotary-wing drones and 2 fixed-wing drones. Ten target points were randomly generated in the mission set, of which 6 were "reconnaissance missions" with loose time window constraints, and 4 were "precision delivery missions" (±2min) with strict time window constraints.

[0172] In 50 independent repeated experiments, regarding feasibility, only 18% of the experiments in control group A had no infeasible paths, control group B improved to 46%, while the proposed invention group achieved 100% feasibility, representing an improvement of 82 percentage points compared to the traditional two-stage method and 54 percentage points compared to the non-feedback joint optimization method. In terms of average total energy consumption of the cluster, control group A was 876 kW·h, control group B was 743 kW·h, and the proposed invention group was only 672 kW·h, a reduction of 23.3% compared to the traditional two-stage method and a further reduction of 9.6% compared to the non-feedback joint optimization method. Regarding on-time task completion rate, control group A was 88%, control group B was 92%, and the proposed invention group reached 97%, representing improvements of 9 and 5 percentage points respectively. In terms of average planning time, control group A was 8.5 seconds, control group B was 14.7 seconds, and the proposed invention group was 12.3 seconds. It is worth noting that although the present invention takes 45% longer than the traditional two-stage method, it actually reduces the overall planning time by 16.3% compared to the joint optimization method without a feedback mechanism, because the feedback mechanism effectively guides the search direction and reduces invalid searches.

[0173] Experimental data shows that, compared to traditional two-stage methods, this invention completely eliminates the fatal flaw of "allocation infeasibility," while significantly reducing energy consumption and improving task completion rate. Compared to joint optimization methods without feedback mechanisms, this invention, through a "feasibility feedback-driven adaptive pruning mechanism for the search space," not only further improves the feasibility of the solution and the quality of task completion, but also enhances optimization efficiency by guiding the search, resulting in a reduction in planning time compared to joint optimization methods without feedback. This fully demonstrates the incremental technical effect brought about by the core feature of "feasibility closed-loop feedback" in this invention—not only solving the feasibility problem, but also improving optimization efficiency and result quality by guiding the search.

[0174] By adopting the above-disclosed technical solution of this invention, the following beneficial effects are obtained:

[0175] By introducing a feasibility feedback mechanism into the joint optimization process of task allocation and trajectory planning, planning individuals that do not meet flight constraints are promptly screened during the search process, and the pheromone concentration is modulated using the verification results to guide the search direction. This makes the planning process more in line with the actual flight conditions of UAVs, solves the problem of infeasible solutions in traditional two-stage methods, and improves optimization efficiency and the quality of planning results.

[0176] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for task and trajectory joint planning of heterogeneous UAV swarm based on feasibility feedback, characterized in that, Includes the following steps: S1. Obtain task set information, drone cluster status information, and task environment information; S2. Construct a multi-dimensional matching degree matrix between drones and tasks based on the task set information, drone cluster status information, and task environment information. S3. Generate an initial planning population based on the multidimensional matching degree matrix, wherein each planning individual is represented by a joint coding structure, which includes task allocation coding and track key point coding. S4. Perform a feasibility closed-loop verification for the planning individuals in the planning population, including: determining the task execution sequence corresponding to each UAV based on the task allocation code; generating a three-dimensional flight track of the corresponding UAV based on the task execution sequence, track key point code, and task environment information; calculating the expected flight energy consumption and expected task completion time of each UAV based on the three-dimensional flight track; comparing the expected flight energy consumption with the remaining battery power constraint of the corresponding UAV, and comparing the expected task completion time with the time window constraint of the corresponding task, to obtain the feasibility verification result of the planning individuals; S5. Based on the feasibility verification results, the planning individuals are screened and adjusted. For planning individuals that do not meet the constraints, fitness penalties or deletion are applied, while planning individuals that meet the constraints are retained. S6. Construct a joint optimization objective function based on the planning individuals that have passed the feasibility verification; S7. The population is iteratively optimized using a joint search strategy that combines genetic evolution operations and pheromone update mechanisms. The feasibility verification results are used to modulate the pheromone concentration of the task allocation code and the track key point code combination. Based on the updated pheromone concentration, the selection, crossover and mutation operations of the planned individuals are performed to generate a new planned population. The method of modulating pheromone concentration using feasibility verification results specifically includes: increasing the corresponding pheromone concentration for task allocation codes and track key point codes that pass feasibility verification; decreasing the corresponding pheromone concentration for task allocation codes and track key point codes that fail feasibility verification; wherein, pheromone concentration serves as heuristic information to participate in the selection operation of planning individuals, guiding the population to converge toward the feasible solution space. S8. When the preset termination conditions are met, output the task allocation and three-dimensional trajectory joint planning scheme that meets the flight constraints.

2. The method of claim 1, wherein, A local repair step is also included between steps S4 and S5: When the expected flight energy consumption of a certain planning individual exceeds the remaining power constraint of the UAV but the excess is less than a preset threshold, a local flight path replanning is triggered for the planning individual. The local flight path is regenerated by adjusting the code of the key points of the flight path, and the feasibility closed-loop verification is re-executed based on the replanning result. The preset threshold is dynamically determined based on the UAV's battery discharge characteristic curve and task priority. For power consumption during the discharge plateau period or high-priority tasks, a larger threshold is used to increase the chance of replanning; for power consumption at the end of the discharge period or low-priority tasks, a smaller threshold is used to strictly screen feasible solutions.

3. The method of claim 1, wherein, The trajectory key point encoding is used to represent the three-dimensional sub-target nodes that guide the UAV's flight. During trajectory generation, the path planner starts from the UAV's current position, passes through the sub-target nodes corresponding to the trajectory key point encoding in sequence, and finally reaches the mission point. By connecting the sub-target nodes in the three-dimensional grid environment, a flyable trajectory that meets obstacle avoidance constraints is generated. The trajectory key point encoding indirectly guides the flight path shape by controlling the spatial distribution of the sub-target nodes.

4. The method of claim 1, wherein, The multidimensional matching degree matrix comprehensively considers at least two of the following factors: the compatibility between the UAV type and the mission type, the distance between the current location of the UAV and the mission point, the matching degree between the UAV's payload capacity and the mission payload requirements, and the matching degree between the UAV's remaining battery power and the estimated flight energy consumption.

5. The method of claim 1, wherein, The joint optimization objective function is used to perform weighted joint optimization of task execution benefits, total flight energy consumption of UAV swarm, and task completion time span, under the premise of satisfying feasibility constraints.

6. The method of claim 1, wherein, The output task allocation and three-dimensional trajectory joint planning scheme is a Pareto optimal solution set composed of multiple planning schemes obtained through a multi-objective optimization algorithm, which are mutually independent in the three objective dimensions of task execution benefits, total energy consumption of UAV swarm flight, and task completion time span. 7.A system for mission and trajectory joint planning of heterogeneous UAV swarm based on feasibility feedback, characterized in that, include: The data acquisition module is used to acquire task set information, UAV cluster status information, and task environment information; The matching degree construction module is used to build a multi-dimensional matching degree matrix between the drone and the mission; The population generation module is used to generate an initial planned population using a joint coding structure. The feasibility verification module is used to generate a three-dimensional flight track based on the task allocation code and the track key point code, and output the feasibility verification results based on the remaining power constraints and the task time window constraints. The joint search module is used to perform fitness adjustment and pheromone concentration update based on the feasibility verification results, and to iteratively generate a new planning population through selection, crossover and mutation operations; wherein, the joint search module uses the feasibility verification results to modulate the pheromone concentration of the combination of task allocation code and track key point code, and guides the population to converge toward the feasible solution space; The command output module is used to output a joint planning scheme for task allocation and three-dimensional trajectory that meets the constraints.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 6.

Images