Multi-unmanned aerial vehicle cooperative path planning method based on complex environment

Abstract

The invention discloses a multi-unmanned aerial vehicle cooperative path planning method based on a complex environment, and the method comprises the following steps: 1, constructing a multi-priority task cooperative scheduling model, and preparing unmanned aerial vehicles and a task data set for model adaptation; step 2, inputting the unmanned aerial vehicle and task parameters into a pre-constructed multi-priority task cooperative scheduling model, and performing initial allocation on tasks according to priorities by the model; 3, collecting state parameters of the unmanned aerial vehicle during task execution in real time, including the current position, the residual load, the residual electric quantity and the distance from the unmanned aerial vehicle to a target point, and determining trigger conditions of material putting, hovering and hovering; 4, completing material delivery, reconnaissance hovering, cooperative following and power supply through a quantitative method of load updating, hovering control and charging scheduling; and 5, according to the task completion progress and the unmanned aerial vehicle resource state, a subsequent task execution plan is adjusted according to a cooperative scheduling principle, and a final task completion result is output.

Description

Technical Field

[0001] The present invention relates to a method for unmanned aerial vehicle (UAV) path planning, in particular to a method for collaborative path planning of multiple UAVs in a complex environment. Background Art

[0002] The application of multi-UAV collaborative path planning technology in military reconnaissance, material delivery and other fields is becoming increasingly widespread. Through multi-UAV collaborative operations, large-scale target coverage and efficient task execution can be achieved, significantly improving operational efficiency in complex environments.

[0003] However, existing technologies still have many key problems in practical applications: in static multi-target coverage scenarios, it is difficult to optimize the total flight path while meeting the drone communication distance constraints and minimum safety spacing constraints, which often leads to unbalanced task allocation or path redundancy; in dynamic environments, when faced with sudden obstacles or new emergency targets, there is a lack of efficient real-time obstacle avoidance and task rescheduling mechanisms, which easily prolongs the task completion time; in multi-priority task scenarios, due to the lack of coordination between drone load, endurance and charging scheduling, problems such as high-priority task delays, untimely drone power warning responses, or low resource utilization often occur.

[0004] Therefore, in order to address the core technical bottlenecks in the above-mentioned complex environment, the development of a multi-UAV collaborative path planning system that takes into account path optimization, dynamic response and resource coordination is of great significance for improving the stability and efficiency of multi-UAV cluster operations. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the purpose of the present invention is to provide a collaborative path planning method for multiple UAVs in a complex environment. The method can achieve efficient coverage of static targets, real-time avoidance of dynamic obstacles, rapid response to emergency tasks, and collaborative resource scheduling of multi-priority tasks, and take into account task coverage, path efficiency and communication stability in low-altitude complex environments.

[0006] To achieve the above objectives, the present invention provides a method for collaborative path planning of multiple UAVs in a complex environment, which is characterized by comprising the following steps: Step 1: Build a multi-priority task collaborative scheduling model and prepare drone and task datasets for model adaptation; Step 2: Input the UAV and mission parameters into a pre-built multi-priority task collaborative scheduling model, and the model will initially allocate tasks according to priority; Step 3: Real-time collection of the drone's status parameters during mission execution, including current position, remaining payload, remaining battery power, and distance to the target point, to determine the triggering conditions for material delivery, hovering, and return; Step 4: Complete material delivery, reconnaissance hovering, coordinated tracking, and battery replenishment through quantitative methods of payload update, hovering control, and charging scheduling; Step 5: According to the task completion progress and the status of the UAV resources, adjust the subsequent task execution plan according to the collaborative scheduling principle and output the final task completion result.

[0007] Preferably, the multi-priority task collaborative scheduling model includes an optimized task allocation module, an optimized dynamic execution module and an optimized resource collaboration module. The optimized task allocation module adopts a combination of bottleneck method, binary method and Hungarian algorithm to perform three-layer priority matching on multi-priority tasks, wherein the task status information of priority 2 and 3 will be transmitted once before matching, and the task status information of the first transmission is named high-priority task status information, and the task status information of the second transmission is medium-priority task status information; The optimized dynamic execution module performs a real-time status information transmission, which is named real-time execution status information; the optimized dynamic execution module includes 4 core functional sub-modules, 1 status monitoring module and a constraint adaptation module.

[0008] Preferably, a progressive constraint adaptation method is introduced in the initial stage of task allocation based on the collaborative path planning method for multiple UAVs in a complex environment to optimize the coordination of the initial path; The optimized resource collaboration module receives high-priority task status information, medium-priority task status information and real-time execution status information, and simultaneously adopts the resource scheduling sub-module in the optimized resource collaboration module to fuse the high-priority task status information, medium-priority task status information and real-time execution status information in the fusion layer, and iteratively optimizes the fused scheduling scheme to form a final path planning result that meets full task coverage, resource adaptation and constraint compliance.

[0009] Preferably, the specific steps of the UAV collaborative scheduling method in step 2 include the following: ①. Determine the variable set: Maximum payload of drone: C={C1,C2...C m}; At t seconds, the drone's current load is: c={c1,c2...c m}; Target point material demand: D={d1,d2...d n}, where d j Represents the jth task point t j material needs; The feasible set of drones at t seconds is: {t 11 ,t12 ...}{t 21 ,t 22 ...}...{t v1 ,t v2 ...}; Among them, {t v1 ,t v2 ...} indicates that the vth drone satisfies c i ≦d j The target point of (1≦j≦n); UAV {u1,u2...u m}The only corresponding assignment is: {t 1y1 ,t 2y2 ...t mym}, where t iyi Indicates drone u i The only target point selected, t iyi ∈ feasible set {t i1 ,t i2 ...}, t iyi Can be nullable (i.e., a target cannot be assigned); ②、Optimization model: Create decision variables:

[0010] According to the decision variables, the objective function is further obtained:

[0011] Constraints: ; ③、Bottleneck model establishment: Optimal allocation: {t 1x1 ,t 2x2 ...t mxm} According to the optimal allocation binary search: define the threshold θ, and find the minimum θ through binary search so that the following allocation scheme satisfies D ij ≤ 0,∀ x ij =1, where D ij It's a drone i With target t j The geometric distance, the search and rescue process meets, The termination conditions are: ; ④. Hungarian method modeling: For binary search, we can get the maximum threshold θ and solve the optimal allocation of drones according to the Hungarian algorithm: {t 1x1 ,t 2x2 ...t mxm}, For the drone u at time t i The distance matrix with the feasible set {tu1, tu2... tu m} is as follows:

[0012] Among them, the targets that do not meet the load requirements are set to inf, where inf represents a very large number, and D ij represents the geometric distance between the drone u i and the target point t j Construct a bipartite graph to solve: G=(U∪T, E), and the edge set is {( E= , i , j )| j ∈ F i ∧ D ij ≤θ}, Find the maximum matching through the Hungarian algorithm M ⊆ E ​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​Indicates drone u j1 The target point, For {t 1x1 ,t 2x2 ...t mxm}\{t j1 x j1 ,t j2 x j2 ,t j3 x j3 ...}, we adopt the "retry" strategy and repeatedly solve until a feasible allocation is found, so that |M| = min(mk min ,n), Among them, k min is the minimum k=0,1,...m that satisfies the conditions, At this point, the target allocation model is fully established. Through the allocation model, we can finally get the following allocation: mk drones u={u1,u2...u m-k The only optimal target point of} is: t 1, t 2... t m-k}, k drones u={u m-k+1 ,u m-k+2 ,...,u m}'s empty set point is {t m-k+1 ,...t m}=∅.

[0013] Preferably, the specific steps of the material delivery method in step 4 include the following: ①, Release trigger: When the real-time position of the drone d is in line with the target point (p k ,q k )Satisfy distance constraints | p(t)-p |< E delivery ( E delivery =20m), the delivery preparation process is triggered. d (t) is the position of UAV d at time t, ε delivery is a small tolerance value, indicating that the UAV has reached the target point; ②、Load update: After the drone d is launched, the payload of the drone d is updated ld( t *)=max(ld( t ) - qk ,0), where l d (t) represents the current payload of UAV d. After completing the mission, the target point is removed from the set T; ③. Return trigger: If the current payload of all the drones cannot meet the requirements of the next mission∃ (p k ,g k )∈ T d satisfy l d ( t ) < q k , then all drones enter the return state, and the target point is set to b; ④. Return motion: The speed of the drone during the return motion is as follows , where k attract is the attraction coefficient, ε is a minimum value to prevent division by zero; ⑤、Resupply completed: The conditions for drone resupply completion are: p d ( t )- b |< ε ( ε =20m), if the supply is completed, reset the load and exit the return state l d ( t + ) = C d , T d ←Reassigned tasks.

[0014] Preferably, the specific steps of the reconnaissance hovering method in step 4 include the following: ①、Hovering judgment: When the distance between the UAV and the target meets: p d ( t ) - p target |<20 (hover radius), activate the fly-around mode, otherwise keep moving towards the target point; ②、Location update: after time Δt p d ( t +Δ t ) =p d ( t ) +v d ( t +Δ t ) · Δ t .

[0015] Preferably, the specific steps of the collaborative following method in step 4 include the following: ①、Select: For all air target drones u i , select the nearest target drone u j , ensuring min{|| u i - u j ||}, any u j ∈{u1,...,u m-k}, where u i ∈{u m-k+1 ,u m-k+2 ,...,u m},u j ∈{u1,u2...u m-k}; ②、Copy: Copy the nearest drone u j The target is to give the air target drone u i ,Right now t ixi = t jxj ; ③、Clear: When the target t jxj Complete the material delivery and clear the global drone target: { t 1, t 2... t m-k}{ t jxj}.

[0016] Preferably, the specific steps of the power replenishment method in step 4 include the following: ①、Charging warning: When any drone meets When the UAV is charging, it will issue a warning; T safe For safety buffer time (20-30 seconds), D i (t) is the Euclidean distance between UAV i and the origin, T i remain (t) is the remaining flight time of drone i at time t. When the above conditions are triggered, all drones will return to the home position. ②Real-time monitoring alarm: Each drone periodically broadcasts its own status (x i ,y i ,T i remain ), perform global most dangerous calculation, , if T critical <T safe , execute the return command; ③、Coordinated return control: Return trigger: If the current payload of all the drones cannot meet the requirements of the next mission∃ (p k ,g k )∈T d satisfy l d ( t ) < q k , then all drones enter the return state, and the target point is set to b; return motion: the speed of the drone during the return motion is as follows , where k attract is the attraction coefficient, ε is a minimum value to prevent division by zero; ④. Latest return calculation: For each drone i, the shortest time required to return is: t i return =D i / v max To ensure safety, the following conditions must be met: .

[0017] Preferably, the calculation method for location update in step 4 includes the following specific steps: ①、Speed ​​update: v d ( t +Δ t ) = v d (t)+ γ ( t ), The control function γ(t) is defined as:

[0018] ②. Attraction

[0019] ③ Speed ​​Constraint: .

[0020] Preferably, the drone resource status in step five includes payload and battery power.

[0021] The technical effects of the present invention are as follows: 1. The present invention improves the multi-UAV collaborative path planning model, including an optimized task allocation module, a dynamic execution module, and a resource collaboration module, which can adapt to more complex task scenarios, including multi-priority task scheduling, dynamic obstacle avoidance, and resource constraint adaptation. This enables the model to optimize the path planning logic from multiple dimensions such as task priority, real-time status, and resource adaptation, thereby improving the efficiency and stability of collaborative path planning.

[0022] 2. The present invention also provides a path planning strategy based on a hierarchical optimization framework, and combined with a multi-UAV collaborative path planning model, it can improve the adaptability to complex environments, enabling it to achieve efficient coverage of static targets, real-time response to dynamic scenes, and orderly execution of multi-priority tasks. It not only improves the task completion rate and resource utilization of multi-UAV collaborative operations, but also can adapt to various complex low-altitude environments such as military reconnaissance and emergency rescue, enhancing the practical application value of multi-UAV cluster operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The generated UAV and mission location map; Figure 2 Matching graph for the Hungarian algorithm; Figure 3 This is the rendering of the return and resupply execution; Figure 4 This is the hover effect picture; Figure 5 This is the effect picture of returning home due to insufficient battery; Figure 6 This is the task recovery effect diagram after charging; Figure 7 This is a flow chart of the collaborative path planning method for multiple UAVs in a complex environment; Figure 8 This is a diagram of the allocation model based on the improved bottleneck method; Figure 9 For collaborative following model diagram; Figure 10 Supplement model diagrams for delivery and return trips; Figure 11 It is a hover model diagram; Figure 12 Diagram of the early warning model for the synchronous return and charging of a drone swarm. DETAILED DESCRIPTION

[0024] To further illustrate the technical implementation and effects of the present invention in multi-priority tasks and charging scheduling scenarios, the specific implementation method, operation process and key parameters of the multi-UAV collaborative path planning strategy are described in detail below, combined with simulation data, model steps and experimental results.

[0025] Reference Figure 1-12 As shown, the present invention provides a method for collaborative path planning of multiple UAVs in a complex environment, comprising the following steps: Step 1: Build a multi-priority task collaborative scheduling model and prepare drone and task datasets for model adaptation; The multi-priority task collaborative scheduling model includes an optimized task allocation module, an optimized dynamic execution module and an optimized resource collaboration module. The optimized task allocation module adopts a combination structure of "bottleneck method + dichotomy + Hungarian algorithm" to perform three-level priority matching processing on multi-priority tasks, wherein the task status information of priority 2 and 3 will be transmitted once before matching. The task status information transmitted for the first time is named high-priority task status information, and the task status information transmitted for the second time is medium-priority task status information; The optimized dynamic execution module performs a real-time status information transmission, which is named real-time execution status information; the optimized dynamic execution module includes 4 core functional sub-modules, 1 status monitoring module and a constraint adaptation module.

[0026] Step 2: Input the UAV and mission parameters into a pre-built multi-priority task collaborative scheduling model, and the model will initially allocate tasks according to priority; The specific steps of the UAV collaborative scheduling method include the following: ①. Determine the variable set: Maximum payload of drone: C={C1,C2...C m}; At t seconds, the drone's current load is: c={c1,c2...c m}; Target point material demand: D={d1,d2...d n}, where d j Represents the jth task point t j material needs; The feasible set of drones at t seconds is: {t 11 ,t 12 ...}{t 21 ,t 22 ...}...{t v1 ,t v2 ...}; Among them, {t v1 ,t v2 ...} indicates that the vth drone satisfies c i ≦d j The target point of (1≦j≦n); UAV {u1,u2...u m}The only corresponding assignment is: {t 1y1 ,t 2y2 ...t mym}, where t iyi Indicates drone u i The only target point selected, t iyi ∈ feasible set {t i1 ,t i2 ...}, tiyi Can be nullable (i.e., a target cannot be assigned); ②、Optimization model: Create decision variables:

[0027] According to the decision variables, the objective function is further obtained:

[0028] Constraints: ; ③、Bottleneck model establishment: Optimal allocation: {t 1x1 ,t 2x2 ...t mxm}

[0029] According to the optimal allocation binary search: define the threshold θ, and find the minimum θ through binary search so that the following allocation scheme satisfies D ij ≤ 0,∀ x ij =1, where D ij It's a drone i With target t j The geometric distance, the search and rescue process meets, The termination conditions are: ; ④. Hungarian method modeling: For binary search, we can get the maximum threshold θ and solve the optimal allocation of drones according to the Hungarian algorithm: {t 1x1 ,t 2x2 ...t mxm}, For drone u at t seconds i and the feasible set {tu1,tu2...tu m The distance matrix of} is:

[0030] Among them, the target that does not meet the load requirement is set to inf, inf represents a very large number, D ij Indicates drone u i and the target point t j Geometric distance, Construct a bipartite graph to solve: G=(U∪T,E), the edge set is E= {( i , j )| j ∈ F i ∧ D ij ≤θ}, Finding the maximum matching through the Hungarian algorithm M ⊆ E Satisfy |M| = min(m, n), where m represents the number of UAVs, n represents the number of target points. When m > n, n UAVs get the optimal unique allocation, and the remaining m - n UAVs are ∅; when m < n, m UAVs all get the optimal unique allocation; ⑤. Rejection + retry strategy: For the k UAVs in the subset {t j1 x j1 , t j2 x j2 , t j3 x j3 ...} ∈ {t 1x1 , t 2x2 ... t mxm}, we use the "rejection" strategy to make the target points of the k UAVs {u j1 , u j2 , u j3 ... u jk}: { t j1 x j1 , t j2 x j2 ,..., t jk x jk} = ∅, where t j1 x j1 represents the target point of UAV u j1 . For {t 1x1 , t 2x2 ... t mxm}\{t j1 x j1 , t j2 x j2 , t j3 x j3 ...}, we adopt the "retry" strategy and repeat the solution until a feasible allocation is found, so that |M| = min(m - k min , n) in the formula, where k min is the minimum k that satisfies the condition, k = 0, 1,... m, At this point, the target allocation model is completely established. Through the allocation model, finally we can get the following allocation: m - k UAVs u = {u1, u2... u m-kThe only optimal target point of} is: t 1, t 2... t m-k}, k drones u={u m-k+1 ,u m-k+2 ,...,u m}'s empty set point is {t m-k+1 ,...t m}=∅.

[0031] Step 3: Real-time collection of the drone's status parameters during mission execution, including current position, remaining payload, remaining battery power, and distance to the target point, to determine the triggering conditions for material delivery, hovering, and return; Step 4: Complete material delivery, reconnaissance hovering, coordinated tracking, and battery replenishment through quantitative methods of payload update, hovering control, and charging scheduling; The specific steps of the material delivery method in step 4 above include the following: ①, Release trigger: When the real-time position of the drone d is in line with the target point (p k ,q k )Satisfy distance constraints | p(t)-p |< E delivery ( E delivery =20m), the delivery preparation process is triggered. d (t) is the position of UAV d at time t, ε delivery is a small tolerance value, indicating that the UAV has reached the target point; ②、Load update: After the drone d is launched, the payload of the drone d is updated ld( t *)=max(ld( t ) - qk ,0), where l d (t) represents the current payload of UAV d. After completing the mission, the target point is removed from the set T; ③. Return trigger: If the current payload of all the drones cannot meet the requirements of the next mission∃ (p k ,g k )∈ T d satisfy l d ( t ) < q k , then all drones enter the return state, and the target point is set to b; ④. Return motion: The speed of the drone during the return motion is as follows , where k attract is the attraction coefficient, ε is a minimum value to prevent division by zero; ⑤、Resupply completed: The conditions for drone resupply completion are: p d ( t )- b |< ε ( ε =20m), if the supply is completed, reset the load and exit the return state l d ( t + )= C d , T d ←Reassigned tasks.

[0032] The reconnaissance hovering method in step 4 above (refer to Figure 11 The specific steps include: ①、Hovering judgment: When the distance between the UAV and the target meets: p d ( t ) - p target |<20 (hover radius), activate the fly-around mode, otherwise keep moving towards the target point; ②、Location update: after time Δt p d ( t +Δ t ) =p d ( t ) +v d ( t +Δ t ) · Δ t .

[0033] The collaborative following method in step 4 above (refer to Figure 9 The specific steps include: ①、Select: For all air target drones u i , select the nearest target drone u j , ensuring min{|| u i - u j ||}, any u j ∈{u1,...,u m-k}, where u i ∈{um-k+1 ,u m-k+2 ,...,u m},u j ∈{u1,u2...u m-k}; ②、Copy: Copy the nearest drone u j The target is to give the air target drone u i ,Right now t ixi = t jxj ; ③、Clear: When the target t jxj Complete the material delivery and clear the global drone target: { t 1, t 2... t m-k}{ t jxj}.

[0034] The specific steps of the power replenishment method in step 4 include the following: ①、Charging warning: refer to Figure 12 , when any drone satisfies When the UAV is charging, it will issue a warning; T safe For safety buffer time (20-30 seconds), D i (t) is the Euclidean distance between UAV i and the origin, T i remain (t) is the remaining flight time of drone i at time t. When the above conditions are triggered, all drones will return to the home position. ②Real-time monitoring alarm: Each drone periodically broadcasts its own status (x i ,y i ,T i remain ), perform global most dangerous calculation, , if T critical <T safe , execute the return command; ③、Coordinated return control: Return trigger: If the current payload of all the drones cannot meet the requirements of the next mission∃ (p k ,g k )∈T d satisfy l d ( t ) < q k, then all drones enter the return state, and the target point is set to b; return motion: the speed of the drone during the return motion is as follows , where k attract is the attraction coefficient, ε is a minimum value to prevent division by zero; ④. Latest return calculation: For each drone i, the shortest time required to return is: t i return =D i / v max To ensure safety, the following conditions must be met: .

[0035] The specific steps of the calculation method for the location update in the above step 4 are as follows: ①、Speed ​​update: v d ( t +Δ t ) = v d (t)+ γ ( t ), The control function γ(t) is defined as:

[0036] ②. Attraction

[0037] ③ Speed ​​Constraint: .

[0038] Step 5: Based on the task completion progress and the UAV resource status, which includes payload and power, the subsequent task execution plan is adjusted according to the collaborative scheduling principle, and the final task completion result is output.

[0039] The aforementioned collaborative path planning method for multiple UAVs in complex environments introduces a “progressive constraint adaptation” method in the initial stage of task allocation to optimize the coordination of the initial paths. The optimized resource collaboration module receives high-priority task status information, medium-priority task status information and real-time execution status information, and simultaneously adopts the resource scheduling sub-module in the optimized resource collaboration module to fuse the high-priority task status information, medium-priority task status information and real-time execution status information in the fusion layer, and iteratively optimizes the fused scheduling scheme to form a final path planning result that meets "full task coverage, resource adaptation, and constraint compliance".

[0040] Reference Figure 1-12As shown, the present invention provides a multi-UAV collaborative path planning method based on a complex environment. The multi-UAV collaborative path planning system mainly includes the following modules: an allocation model based on the improved bottleneck method, a collaborative following model, a delivery and return replenishment model, a hovering model, and a UAV swarm synchronous return charging warning model; the implementation ideas of the system are divided into the following four modules, and the optimization of these four modules is proved respectively. The following is a layer-by-layer deduction based on the core formulas in the specification and the embodiments to reveal the inherent mathematical principles of the technical effect: 1. Proof of Optimization of Task Assignment Module 1. Optimality of the Hungarian Algorithm (for results refer to Figure 2 ) Original formula:

[0041] Derivation process: Step 1: Construct a bipartite graph (G=( U ∪ T , E ), edge weight ( w ij = - D ij ) (The minimization problem is transformed into maximizing negative weight matching); Step 2: According to König's theorem, the maximum weight matching of a bipartite graph is equivalent to the sum of the weights of the minimum vertex cover; Step 3: The augmented path search of the Hungarian algorithm is guaranteed to find the global optimal match, and its complexity is (O( mn 2 )).

[0042] Technical effects: The theoretical minimum value of the total distance of task allocation is guaranteed by the duality of linear programming. In Example 1, the total flight distance of 4500m (A+B+C) has reached the lower bound.

[0043] 2. Convergence of binary threshold search Original formula: , termination condition θ low ≥ θ high ,

[0044] Derivation process: Step 1: Monotonicity proof. For any θ1<θ2, if θ1 is feasible, then θ2 must be feasible (because D ij ≤ θ1⇒ D ij ≤θ2); Step 2: Each iteration reduces the search space by half, and the number of convergence steps is

[0045] Step 3: Finally, θ satisfies ∀ x ij =1, D ij ≤ θ, and there is no smaller θ´ that satisfies the condition.

[0046] Technical effects: In Example 1 θ =500 m The solution only requires 7 iterations, which improves the efficiency by 57%.

[0047] 2. Proof of Stability of the Dynamic Obstacle Avoidance Module 1. Lyapunov function of virtual force model Original formula:

[0048] Derivation process: Step 1: Define the Lyapunov function .

[0049] Step 2: Derivative .

[0050] Step 3: Substitute into the speed formula and prove when P d ≠ b .

[0051] Technical effects: The drone must converge to the target point ( Figure 3 Return trajectory verification), and steady-state error | p d - b |< ε =20m.

[0052] 3. Robustness Proof of Resource Collaboration Module 1. Markov property of power warning Original formula:

[0053] Derivation process: Step 1: Model as a continuous-time Markov chain, state (s=( T remain , D i )); Step 2: Define the dangerous state set; Step 3: Through the transfer rate matrix Q Calculate the absorption probability.

[0054] Technical effects: when T safe = 30s, the warning trigger probability in Example 4 is > 99.9%, refer to Figure 5 No depleted battery shown.

[0055] 2. Existence of Nash Equilibrium in Load Balancing Original formula: l d ( t * )=max( l d ( t ) - q k ,0)

[0056] Derivation process: Step 1: Define the utility function ( U i = C i - l d ( t )); Step 2: Prove that the game Γ=〈{ u i},{ U i},{ C i}〉is a potential game, the potential function (Φ = ∑ U i ); Step 3: There must be a pure strategy Nash equilibrium in a potential game (Monderer & Shapley, 1996).

[0057] Technical effects: In Example 2, the load distribution variance after the return decision was reduced from 12.3 to 2.1 ( Figure 3 ).

[0058] 4. Emergent Proof of System-Level Effectiveness 1. Queuing Theory Upper Bound of Task Completion Time Correlation formula: t i return ≤ T i remain - T safe

[0059] Derivation process: Step 1: Model as M / M / m queue, task arrival rate λ, service rate μ=1 / E[ D i / V max ]; Step 2: Apply Kingman's formula, average waiting time:

[0060] Step 3: Calculate the total time Technical effects: The error between the total time of Example 1-4 (656s) and the theoretical value (623s) is <5% Reference Figures 1-12 As shown, the present invention relates to a multi-UAV collaborative path planning system based on a complex environment. A data set is produced to verify whether the UAV path planning is accurate. The data set includes two parts: UAV parameters and mission parameters. The verification is as follows: Example 1: Multi-priority task allocation and basic scheduling Scenario setting: Three drones need to perform 10 tasks (including three priority 1 emergency delivery tasks, five priority 2 normal delivery tasks, and two priority 3 reconnaissance hovering tasks). They need to be executed in order of priority and meet the load, communication distance (≤1000m) and power constraints; Data preparation: The drone parameters and mission parameters are randomly generated through computer seeds. The drone parameters include the number of drones, maximum load, maximum endurance and reconnaissance hovering time; the mission parameters include mission priority, mission coordinates, material weight or hovering requirements, and the target points and drones are generated. Figure 1 shown.

[0061] Drone parameters:

[0062] Task parameters:

[0063] Implementation steps: 1. Task Prioritization Execute in the order of priority 1 → 2 → 3, giving priority to tasks with priority 1; 2. Bottleneck task allocation (refer to Figure 8 ) Objective: Under the constraints of UAV payload, find the optimal allocation that minimizes the maximum mission distance using the bisection method combined with the Hungarian algorithm. Priority 1 Task Assignment: Feasible set screening: T1 (7kg) can be performed by A (8kg), B (14kg), and C (15kg); T2 (14kg) and T3 (14kg) can be performed by B (14kg) and C (15kg); Binary search threshold θ: The minimum θ = 500m is determined by the distance matrix (Euclidean distance between drones and tasks) to ensure that all tasks can be assigned; Hungarian algorithm matching: A→T1 (distance 450m), B→T3 (distance 600m), C→T2 (distance 550m); 3. Cooperative following and communication constraint maintenance (refer to Figure 9 ) All drones have the same initial position and maintain a distance between them (50m≤d≤1000m) through a virtual force model during mission execution: Attraction: points to the target point (e.g. A is attracted by T1, the direction is T1 - A's current position); Repulsive force: Avoid the distance between drones being less than 50m (for example, when the distance between A and B is 40m, repulsive force will push them apart).

[0064] 4. Verification Priority 1 task completion time: A completes T1 at step=86, B completes T3 at step=146, and C completes T2 at step=110; the effect is as follows Figure 2 shown.

[0065] Total flight distance: A = 1200m, B = 1800m, C = 1500m, all meeting the communication constraints (maximum distance 900m < 1000m).

[0066] Example 2: Material delivery and return supply scheduling (based on the delivery and return replenishment model, refer to Figure 10 ) Scenario setting: After completing the priority 1 task, the remaining payload of the drone is insufficient to execute the priority 2 task (for example, A has 1kg remaining and cannot execute the 8kg requirement of T6), and a return to resupply is triggered.

[0067] Data preparation: The remaining states of the drone are: A (load 1kg), B (load 0kg), C (load 1kg); Priority 2 task requirements: T4 (14kg), T5 (5kg), T6 (8kg), T7 (9kg), T8 (12kg).

[0068] Implementation steps: 1. Delivery completion judgment When the distance between the drone and the target point is less than 20m, the release is triggered and the payload is updated: After A releases T1, the load changes from 8kg to 1kg; After B puts T3, the load changes from 14kg to 0kg; After C releases T2, the load changes from 15kg to 1kg.

[0069] 2. Return trigger condition detection Check whether the remaining load meets the requirements of the next task: A (1kg) < T6 (8kg), B (0kg) < T4 (14kg), C (1kg) < T7 (9kg), triggering a full return.

[0070] 3. Return and resupply execution Return target: origin (0,0), guided by the attraction model (speed is proportional to the distance from the origin); Resupply completion conditions: The drone reaches the origin and the distance is less than 20m, and the payload is reset to the maximum (A=8kg, B=14kg, C=15kg); Time consumption: A reaches the origin in step = 212, B arrives in step = 248, and C arrives in step = 219. The effect is as follows Figure 3 .

[0071] 4. Secondary task allocation and execution Priority 2 task assignment: A→T5 (5kg), B→T7 (9kg), C→T8 (12kg); Completion time: A completes T5 at step=295, B completes T7 at step=305, and C completes T8 at step=305.

[0072] Verification results: Total return time: 38 steps (from completion of mission 1 to second departure); The total flight distance of the second launch is: A=900m, B=1100m, C=1300m, all of which meet the communication constraints.

[0073] Example 3: Hovering reconnaissance mission and coordinated scheduling (based on the hovering model, refer to Figure 4 、 Figure 11 ) Scenario setting: After completing the priority 2 mission, execute the priority 3 reconnaissance mission (T9 requires hovering for 40 seconds, T10 requires hovering for 25 seconds), and the hovering time constraints and communication distance must be met.

[0074] Data preparation: Drone status: A (3kg payload, 500s remaining flight time), B (5kg payload, 450s remaining flight time), C (3kg payload, 400s remaining flight time); Reconnaissance mission: T9 (hover 40 seconds), T10 (hover 25 seconds).

[0075] Implementation steps: 1. Hover Task Assignment Based on the upper limit of hovering time, A (60s) can execute T9 (40s), B (30s) can execute T10 (25s), and C is a collaborative follower drone; Path planning: A from the current position (701.72, 83.55) → T9, B from (280.72, 744.17) → T10.

[0076] 2. Hover Mode Activation and Execution Activation conditions: The drone reaches the vicinity of the target point (distance < 20m) and enters hovering. Attraction: Point to the center of the target point (maintain the flying radius); Speed ​​constraint: When hovering, the speed should be ≤10m / s to avoid deviation; Time count: A starts hovering at step = 613 and continues for 40 steps; B starts hovering at step = 631 and continues for 25 steps.

[0077] 3. Collaborative follow-up maintenance C follows B, which is closest to it, and maintains the distance (800m < 1000m) through virtual force without interfering with the main task.

[0078] 4. Verification Hover completion time: A completes T9 at step=653, and B completes T10 at step=656; Communication stability during hovering: The maximum distance between A and B is 950m, which meets the constraints. The overall effect is as follows: Figure 4 .

[0079] Example 4: Charging warning and synchronous return scheduling (based on the synchronous return charging warning model, refer to Figure 12 ) Scenario setting: During mission execution, the drone's battery power is consumed as the flight time increases. It is necessary to monitor the battery power in real time and return to the base for charging when an alarm is triggered.

[0080] Data preparation: Battery life parameters: A has 200s remaining, B has 180s remaining, and C has 150s remaining. Safety parameters: safety buffer time Tsafe = 30s, average distance between the drone's current position and the origin is 800m, and maximum speed is 50m / s.

[0081] Implementation steps: 1. Real-time power monitoring The drone status is collected every step (1s): remaining flight time, and the distance Di between the current position and the origin.

[0082] 2. Determine the trigger conditions for warnings Based on the formula: Remaining battery life

[0083] C of D i =800m, return time 800 / 50=16s, Tremain=150s ≤ 16+30=46s? Not satisfied; When step = 375, C's remaining endurance = 45s, 45 ≤ 16 + 30 = 46 → triggering an alert.

[0084] 3. Synchronous return execution All drones abandon their current mission and return to their starting point; the effect is as follows Figure 5 Charging time: 60 steps (from step=409 to 469), power reset to maximum (A=600s, B=500s, C=450s).

[0085] 4. Mission resumes after charging Remaining tasks: T4 (14kg), T6 (8kg); Assignment: B→T4, A→T6, completion time: B at step=550, A at step=594, the effect is as follows Figure 6 .

[0086] Verification results: Timeliness of warning triggering: All drones return home safely before battery runs out (C remaining 45s>16s); Task completion time after charging: total time taken is 656 steps, which meets the overall scheduling requirements.

[0087] Summary of implementation effects 1. Constraint satisfaction: In all embodiments, the drones meet constraints such as communication distance, payload, and power consumption; 2. Efficiency: The total completion time for multi-priority tasks was 656 seconds. Path planning and scheduling strategies effectively balanced task priorities and resource utilization. 3. Robustness: Through mechanisms such as return-to-home replenishment and charging warnings, dynamic problems such as insufficient payload and insufficient battery can be addressed to ensure mission continuity.

[0088] The above embodiments completely cover the entire process of multi-priority task allocation, material delivery, hovering reconnaissance and charging scheduling, verifying the feasibility and effectiveness of the model of the present invention in complex environments.

[0089] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for collaborative path planning of multiple UAVs in a complex environment, characterized by: Including the following steps: Step 1: Construct a multi-priority task collaborative scheduling model, and prepare the UAVs and task datasets for model adaptation; Step 2: Input the UAVs and task parameters into the pre-constructed multi-priority task collaborative scheduling model, and the model makes an initial allocation of tasks according to priorities; Step 3: Real-time collect the state parameters of the UAVs during task execution, including the current position, remaining payload, remaining power, and distance to the target point, and determine the triggering conditions for material delivery, hovering, and returning; Step 4: Through the quantification methods of payload update, hovering control, and charging scheduling, complete material delivery, reconnaissance hovering, collaborative following, and power supply; Step 5: According to the task completion progress and the UAV resource status, adjust the subsequent task execution plan based on the collaborative scheduling principle, and output the final task completion result.

2. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 1 is characterized by: The multi-priority task collaborative scheduling model includes an optimized task allocation module, an optimized dynamic execution module, and an optimized resource collaboration module. The optimized task allocation module adopts a combined structure of the bottleneck method, the dichotomy method, and the Hungarian algorithm to perform three-layer priority matching processing on multi-priority tasks. Before the matching of priority 2 and 3 tasks, a task status information transmission will be carried out respectively. The task status information transmitted for the first time is named high-priority task status information, and the task status information transmitted for the second time is medium-priority task status information; The optimized dynamic execution module performs a real-time status information transmission, named real-time execution status information; the optimized dynamic execution module includes 4 core functional sub-modules, 1 status monitoring module, and a constraint adaptation module.

3. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 2 is characterized by: In the initial stage of task allocation of the multi-UAV collaborative path planning method in a complex environment, a progressive constraint adaptation method is introduced to optimize the coordination of the initial path; The optimized resource collaboration module receives high-priority task status information, medium-priority task status information, and real-time execution status information, and at the same time adopts the resource scheduling sub-module in the optimized resource collaboration module to fuse the high-priority task status information, medium-priority task status information, and real-time execution status information in the fusion layer, and iteratively optimize the fused scheduling scheme to form a final path planning result that meets task full coverage, resource adaptation, and constraint compliance.

4. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 1, 2 or 3, characterized in that: The specific steps of the UAV collaborative scheduling method in Step 2 are as follows: ①. Determine the variable set: Maximum payload of drone: C={C1,C2...C m }; At t seconds, the drone's current load is: c={c1,c2...c m }; Target point material demand: D={d1,d2...d n }, where d j Represents the jth task point t j material needs; The feasible set of drones at t seconds is: {t 11 ,t 12 ...}{t 21 ,t 22 ...}...{t v1 ,t v2 ...}; Among them, {t v1 ,t v2 ...} indicates that the vth drone satisfies c i ≦d j The target point of (1≦j≦n); UAV {u1,u2...u m }The only corresponding assignment is: {t 1y1 ,t 2y2 ...t mym }, where t iyi Indicates drone u i The only target point selected, t iyi ∈ feasible set {t i1 ,t i2 ...}, t iyi Can be nullable (i.e., a target cannot be assigned); ②. Optimize the model: Create decision variables: According to the decision variables, the objective function is further obtained: Constraints: ; ③. Establish the bottleneck method model: Optimal allocation: {t 1x1 ,t 2x2 ...t mxm } According to the optimal allocation binary search: define the threshold θ, and find the minimum θ through binary search so that the following allocation scheme satisfies D ij ≤ 0,∀ x ij =1, where D ij It's a drone i With target t j The geometric distance, the search and rescue process meets, The termination conditions are: ; ④. Model the Hungarian method: For binary search, we can get the maximum threshold θ and solve the optimal allocation of drones according to the Hungarian algorithm: {t 1x1 ,t 2x2 ...t mxm }, For drone u at t seconds i and the feasible set {tu1,tu2...tu m The distance matrix of} is: Among them, the target that does not meet the load requirement is set to inf, inf represents a very large number, D ij Indicates drone u i and the target point t j Geometric distance, Construct a bipartite graph to solve: G=(U∪T,E), the edge set is E= {( i , j )| j ∈ F i ∧ D ij ≤θ}, Finding the maximum matching using the Hungarian algorithm M ⊆ E Satisfies |M|= min(m,n), Where m represents the number of UAVs, n represents the number of target points. When m>n, n UAVs get the optimal and unique allocation, and the remaining m - n UAVs are ∅; when m<n, all m UAVs get the optimal and unique allocation; ⑤. Rejection + retry strategy: For the subset {t j1 x j1 ,t j2 x j2 ,t j3 x j3 ...}∈{t 1x1 ,t 2x2 ...t mxm } of k drones, we use the "elimination" strategy to make k drones {u j1 ,u j2 ,u j3 ...u jk }Target point: {t j1 x j1 ,t j2 x j2 ,...,t jk x jk }=∅ , Among them, t j1 x j1 Indicates drone u j1 The target point, For {t 1x1 ,t 2x2 ...t mxm }\{t j1 x j1 ,t j2 x j2 ,t j3 x j3 ...}, we adopt the "retry" strategy and repeatedly solve until a feasible allocation is found, so that |M| = min(mk min ,n), Among them, k min is the minimum k=0,1,...m that satisfies the conditions, So far, the target allocation model is completely established. Through the allocation model, we can finally get the following allocation: mk drones u={u1,u2...u m-k The only optimal target point is: {t1,t2...t m-k }, k drones u={u m-k+1 ,u m-k+2 ,...,u m }'s empty set point is {t m-k+1 ,...t m }=∅.

5. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 1, 2 or 3, characterized in that: The specific steps of the material delivery method in Step 4 are as follows: ①, Release trigger: When the real-time position of the drone d is in line with the target point (p k ,q k )Satisfy distance constraints | p(t)-p |< E delivery ( E delivery =20m), the delivery preparation process is triggered; where p d (t) is the position of UAV d at time t, ε delivery is a small tolerance value, indicating that the UAV has reached the target point; ②、Load update: After the drone d is launched, the payload of the drone d is updated ld( t *)=max(ld( t )- qk ,0), where l d (t) represents the current payload of UAV d. After completing the mission, the target point is removed from the set T; ③. Return trigger: If the current payload of all the drones cannot meet the requirements of the next mission∃ (p k ,g k )∈T d satisfy l d ( t ) < q k , then all drones enter the return state, and the target point is set to b; ④. Return motion: The speed of the drone during the return motion is as follows , where k attract is the attraction coefficient, ε is a minimum value to prevent division by zero; ⑤、Resupply completed: The conditions for drone resupply completion are: p d ( t )- b |< ε ( ε =20m), if the supply is completed, reset the load and exit the return state l d ( t + )= C d , T d ←Reassigned tasks.

6. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 1, 2 or 3, characterized in that: The specific steps of the reconnaissance hovering method in Step 4 are as follows: ①、Hovering judgment: When the distance between the UAV and the target meets: p d ( t ) - p target |<20 (hover radius), activate the fly-around mode, otherwise keep moving towards the target point; ②、Location update: after time Δt p d ( t +Δ t ) =p d ( t ) +v d ( t +Δ t ) · Δ t。 7. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 1, 2 or 3, characterized in that: The specific steps of the collaborative following method in Step 4 are as follows: ①、Select: For all air target drones u i , select the nearest target drone u j , ensuring min{|| u i - u j ||}, any u j ∈{u1,...,u m-k }, where u i ∈{u m-k+1 ,u m-k+2 ,...,u m },u j ∈{u1,u2...u m-k }; ②、Copy: Copy the nearest drone u j The target is to give the air target drone u i ,Right now t ixi = t jxj ; ③、Clear: When the target t jxj Complete the material delivery and clear the global drone target: { t 1, t 2... t m-k }{ t jxj }.

8. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 1, 2 or 3, characterized in that: The specific steps of the power replenishment method in step 4 include the following: ①、Charging warning: When any drone meets When the UAV is charging, it will issue a warning; T safe For safety buffer time (20-30 seconds), D i (t) is the Euclidean distance between UAV i and the origin, T i remain (t) is the remaining flight time of drone i at time t. When the above conditions are triggered, all drones will return to the home position. ②Real-time monitoring alarm: Each drone periodically broadcasts its own status (x i ,y i ,T i remain ), perform global most dangerous calculation, , if T critical <T safe , execute the return command; ③. Coordinated return control: refer to steps ③ and ④ of claim 5, which are exactly the same as the coordinated return for material delivery and replenishment; ④. Latest return calculation: For each drone i, the shortest time required to return is: t i return =D i / v max To ensure safety, the following conditions must be met: .

9. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 1, 2 or 3, characterized in that: The specific steps of the calculation method for the location update in step 4 are as follows: ①、Speed ​​update: v d ( t +Δ t ) = v d (t)+ γ ( t ) The control function γ(t) is defined as: ②. Attraction ③ Speed ​​Constraint: .

10. The method for collaborative path planning of multiple UAVs in a complex environment according to claim 1, 2 or 3, characterized in that: The drone resource status in step 5 includes payload and battery power.

Images