Multi-robot dog cooperative task allocation method, device and equipment and storage medium

Abstract

The invention discloses a multi-robot-dog cooperative task allocation method, device and equipment and a storage medium, and the method comprises the steps: constructing a particle neighborhood between robot dogs through a current time-varying communication topological graph, calculating a topological guidance center, and dynamically adjusting an inertia weight according to the particle neighborhood and the topological guidance center; updating the particle speed and the particle position according to the adjusted inertia weight, and executing hierarchical constraint processing on the updated particle position; based on particle positions after constraint processing, balancing a multi-objective function by using a dynamic weight aggregation mechanism, monitoring an environment state vector in real time to trigger parameter adaptive adjustment, and activating a hybrid local search strategy during convergence stagnation to control each robot dog to search, so that the feasibility of a task allocation scheme can be guaranteed, and the task allocation efficiency is improved. The efficiency, scheme feasibility and environmental adaptability of multi-robot-dog cooperative task allocation are remarkably improved, the optimization quality of the task allocation scheme is improved, and the speed and efficiency of multi-robot-dog cooperative task allocation are improved.

Description

Technical Field

[0001] This invention relates to the field of swarm intelligence optimization and robot collaborative control technology, and in particular to a method, apparatus, device and storage medium for multi-robot collaborative task allocation. Background Technology

[0002] Particle Swarm Optimization (PSO) algorithms are widely used in optimization problems due to their advantages such as simple implementation, high computational efficiency, and fast convergence speed. However, in complex scenarios involving heterogeneous multi-machine dog cooperative task allocation, the classic PSO algorithm reveals the following significant limitations: 1. Communication topology neglect: Traditional PSO does not explicitly consider the communication topology and link quality (such as signal strength, bandwidth or latency) in distributed networks, making it difficult to adapt to dynamically changing heterogeneous network environments, resulting in low communication efficiency or unreliable allocation schemes.

[0003] 2. Insufficient diversity: PSO relies on a single-point learning mechanism of the globally optimal particle, which can easily lead to the population converging to a local optimum too early and lacking a comprehensive exploration of the solution space.

[0004] 3. Parameter staticity: Inertia weights and learning factors are usually fixed values, lacking a mechanism for dynamic adjustment with changes in the environment, which limits the algorithm's adaptability to dynamic scenarios.

[0005] 4. Constraint neglect: The particle position and velocity updates in PSO do not fully incorporate the kinematic constraints of the physical robot (such as velocity limits and rotation angle limits), energy constraints (such as battery capacity), and communication constraints (such as connectivity requirements), resulting in insufficient feasibility of the solution.

[0006] 5. Insufficient trade-offs among multiple objectives: In multi-objective optimization scenarios (such as minimizing total latency, energy consumption, load imbalance, and time limit violations), traditional PSO is difficult to effectively balance the conflicts between objectives, and the optimization results are often biased towards a single objective. Summary of the Invention

[0007] The main objective of this invention is to provide a method, apparatus, device, and storage medium for multi-robot collaborative task allocation, aiming to solve the technical problems in the prior art, such as low task allocation efficiency, infeasibility, and poor environmental adaptability caused by the lack of dynamic perception of communication topology changes, insufficient constraint processing, and rigid multi-objective trade-offs.

[0008] In a first aspect, the present invention provides a method for allocating collaborative tasks among multiple robot dogs, the method comprising the following steps: Based on the current time-varying communication topology, a particle neighborhood is constructed between each robot dog and a topology guidance center is calculated. The inertial weights are dynamically adjusted based on the particle neighborhood and the topology guidance center. The particle velocity and position are updated based on the adjusted inertia weights, and a hierarchical constraint process is applied to the updated particle positions. Based on the particle positions after constraint processing, a dynamic weight aggregation mechanism is used to balance the multi-objective function, and the environmental state vector is monitored in real time to trigger adaptive parameter adjustment. At the same time, when convergence stalls, a hybrid local search strategy is activated to control each robot dog to search.

[0009] Optionally, the step of constructing particle neighborhoods among each robot dog based on the current time-varying communication topology graph and calculating the topology guidance center, and dynamically adjusting the inertial weights based on the particle neighborhoods and the topology guidance center, includes: Obtain the current time-varying communication topology map, and construct a particle neighborhood for each robot dog based on the current time-varying communication topology map; The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score. The inertia weight is dynamically adjusted based on the size of the particle neighborhood and the distribution characteristics of the topological guidance center.

[0010] Optionally, obtaining the current time-varying communication topology map and constructing a particle neighborhood for each robot dog based on the current time-varying communication topology map includes: Obtain the vertex set, communication link edge set, and edge weight communication quality score of the robot dog, and determine the current time-varying communication topology using the following formula:

[0011] in, This is the current time-varying communication topology diagram. For the vertex set of the robot dog, For communication link edge set, Scoring the quality of edge-weighted communication; Real-time collection of status data for each robot dog; dynamic updating of the topology map based on the status data. To select neighboring robot dogs that meet the communication quality score from the set of robot dog vertices for each robot dog, the particle neighborhood is constructed using the following formula:

[0012] in, For the particle neighborhood, For the neighbor's robot dog, For the vertex set of the robot dog, For robot dogs and Inter-link communication quality score, This is a preset scoring threshold.

[0013] Optionally, the step of calculating the topology guidance center using the historical best positions of particles in the particle's neighborhood and communication quality scores includes: The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score, according to the following formula:

[0014] in, As the topology guidance center, For the particle neighborhood, For robot dogs and Inter-link communication quality score, For neighborhood particles Its historical best position.

[0015] Optionally, the step of updating the particle velocity and particle position according to the adjusted inertia weight, and performing hierarchical constraint processing on the updated particle position, includes: The particle velocity is updated according to the adjusted inertia weight using the following formula:

[0016] in, For particles In the The speed of generation For particles In the The inertial weight of the generation, For particles In the The speed of generation Learning factors that control for the influence of individual experience. A random number in the interval [0,1]. For particles The best historical position For particles In the The position of the generation, To control the learning factor of neighborhood guidance influence, A random number in the interval [0,1]. As the topology guidance center, To constrain the penalty coefficient, To constrain the penalties for violations; Perform a position update and obtain the updated particle position using the following formula:

[0017] in, For particles In the The updated position of the generation For particles In the The position of the generation, For particles In the The speed of generation; The feasible region is mapped by hard constraint projection using the following formula:

[0018] in, For particles In the The updated position of the generation For projection operators, This is the feasible region with hard constraints; The penalty term for the degree of violation of soft constraints is applied by the following formula:

[0019] in, To constrain the penalties for violations, For the first The penalty coefficient for a soft constraint, For particles For the The degree of violation of a soft constraint; Heuristic repair strategies are used to reassign tasks, reallocate resources, fine-tune timing, and switch emergency modes for particles that violate constraints.

[0020] Optionally, based on the particle positions after constraint processing, a dynamic weight aggregation mechanism is used to balance the multi-objective function, and the environmental state vector is monitored in real time to trigger adaptive parameter adjustment. Simultaneously, a hybrid local search strategy is activated when convergence stalls to control each robot dog to perform the search, including: Based on the particle positions after constraint processing, the weights of each sub-target are adaptively calculated using the following formula through a dynamic weight aggregation mechanism:

[0021] in, For the first The sub-goals in the The weight of generations For the first The dynamic importance of individual sub-goals; The environment state vector is obtained using the following formula:

[0022] in, For the first The environment state vector of the generation, For changes in network topology, For changes in the task set, This refers to the changes in the robot dog's state. The dynamic importance of each sub-objective is dynamically adjusted based on the environmental state vector and historical optimization performance, thereby balancing the multi-objective function in the following equation:

[0023]

[0024]

[0025]

[0026]

[0027] in, For particles The comprehensive objective function value, The dynamic weights of the sub-objective function for the total time. Let be a sub-objective function of the total time. The dynamic weights of the sub-objective function for total energy consumption. Let be a sub-objective function of total energy consumption. The dynamic weights are the sub-objective functions for load balancing. This is a sub-objective function for load balancing. The dynamic weights of the sub-objective function for time limit violations. For the sub-objective function of time limit violation, To allocate matrix elements, For robot dogs Execute the task Required completion time For robot dogs Execute the task energy consumption For the task For robot dogs Workload weighting It is the variance function. For the task The actual completion time, For the task The deadline This indicates that if the task expires, the expiration time will be returned; otherwise, 0 will be returned. The system monitors the environmental state vector in real time. When the environmental change detection condition is triggered, it dynamically adjusts each parameter and continuously evaluates the convergence state of the algorithm. When convergence stalls, it activates a hybrid local search strategy to control each robot dog to search.

[0028] Optionally, the real-time monitoring of the environmental state vector dynamically adjusts various parameters when the environmental change detection condition is triggered, continuously evaluates the algorithm's convergence state, and activates a hybrid local search strategy to control each robot dog to search when convergence stalls, including: Real-time monitoring of the environmental state vector, and determination of environmental change detection conditions using the following formula:

[0029] in, Let be the environment state vector at the current time t. Let be the environment state vector at the previous time t-1. For dynamic adaptive thresholds; When the environmental mutation detection condition is triggered, the inertia weight, the learning factor that controls the influence of individual experience, the learning factor that controls the influence of neighborhood guidance, and the particle neighborhood size are dynamically adjusted. The algorithm's convergence status is continuously evaluated. When the inverse generational distance or hypervolume index of the Pareto optimal solution set improves below a preset threshold for N consecutive generations, a hybrid local search strategy is adaptively activated. Each robot dog is controlled to search according to the hybrid local search strategy, which includes greedy improvement, simulated annealing, variable neighborhood search, or tabu search.

[0030] Secondly, to achieve the above objectives, the present invention also proposes a multi-robot collaborative task allocation device, the multi-robot collaborative task allocation device comprising: The dynamic adjustment module is used to construct the particle neighborhood between each robot dog based on the current time-varying communication topology map and calculate the topology guidance center, and dynamically adjust the inertial weights based on the particle neighborhood and the topology guidance center; The hierarchical constraint module is used to update the particle velocity and particle position according to the adjusted inertia weight, and to perform hierarchical constraint processing on the updated particle position; The control module is adjusted to balance the multi-objective function using a dynamic weight aggregation mechanism based on the particle positions after constraint processing, and to monitor the environmental state vector in real time to trigger adaptive parameter adjustment. At the same time, when convergence stalls, a hybrid local search strategy is activated to control each robot dog to search.

[0031] Thirdly, to achieve the above objectives, the present invention also proposes a multi-robot collaborative task allocation device, the multi-robot collaborative task allocation device comprising: a memory, a processor, and a multi-robot collaborative task allocation program stored in the memory and executable on the processor, the multi-robot collaborative task allocation program being configured to implement the steps of the multi-robot collaborative task allocation method described above.

[0032] Fourthly, to achieve the above objectives, the present invention also proposes a storage medium storing a multi-machine dog collaborative task allocation program, wherein when the multi-machine dog collaborative task allocation program is executed by a processor, it implements the steps of the multi-machine dog collaborative task allocation method described above.

[0033] The multi-robot collaborative task allocation method proposed in this invention constructs particle neighborhoods among each robot dog using the current time-varying communication topology graph and calculates the topology guidance center. It dynamically adjusts the inertial weights based on these neighborhoods and guidance centers. The method updates particle velocity and position according to the adjusted inertial weights and applies hierarchical constraint processing to the updated particle positions. Based on the constrained particle positions, a dynamic weight aggregation mechanism balances multiple objective functions, and the environmental state vector is monitored in real time to trigger adaptive parameter adjustments. Simultaneously, a hybrid local search strategy is activated when convergence stalls to control each robot dog's search. This ensures the feasibility of the task allocation scheme, significantly improves the efficiency, feasibility, and environmental adaptability of multi-robot collaborative task allocation, enhances the optimization quality of the task allocation scheme, strengthens the algorithm's adaptability to dynamic network topology and communication quality changes, ensures that the allocation scheme meets the hard constraints of kinematics, energy, and communication of the physical robot, improves the algorithm's convergence speed and solution set diversity, avoids getting trapped in local optima, adapts to complex dynamic environments, ensures the robustness and real-time performance of the allocation scheme, and improves the speed and efficiency of multi-robot collaborative task allocation. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the first embodiment of the multi-robot collaborative task allocation method of the present invention; Figure 3 This is a flowchart illustrating the second embodiment of the multi-robot collaborative task allocation method of the present invention; Figure 4 This is a functional block diagram of the first embodiment of the multi-robot collaborative task allocation device of the present invention.

[0035] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0036] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0037] The solution of this invention mainly involves: constructing particle neighborhoods among the robot dogs using the current time-varying communication topology graph and calculating the topology guidance center; dynamically adjusting the inertia weights based on the particle neighborhoods and the topology guidance center; updating the particle velocity and position based on the adjusted inertia weights; performing hierarchical constraint processing on the updated particle positions; using a dynamic weight aggregation mechanism to balance the multi-objective function based on the constrained particle positions; and monitoring the environmental state vector in real time to trigger adaptive parameter adjustment. Simultaneously, a hybrid local search strategy is activated when convergence stalls to control each robot dog's search, ensuring the feasibility of the task allocation scheme and significantly improving multi-robot dog collaboration. The efficiency, feasibility, and environmental adaptability of task allocation are improved, enhancing the optimization quality of task allocation schemes. This strengthens the algorithm's adaptability to changes in dynamic network topology and communication quality, ensuring that the allocation scheme meets the hard constraints of kinematics, energy, and communication of the physical robot. It also improves the convergence speed and solution set diversity of the algorithm, avoids getting trapped in local optima, and can adapt to complex dynamic environments. This ensures the robustness and real-time performance of the allocation scheme, improves the speed and efficiency of multi-robot collaborative task allocation, and solves the technical problems in existing technologies that lead to low task allocation efficiency, infeasibility, and poor environmental adaptability due to the lack of dynamic perception of communication topology changes, insufficient constraint processing, and rigidity in multi-objective trade-offs.

[0038] Reference Figure 1 , Figure 1 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of the present invention.

[0039] like Figure 1 As shown, the device may include: a processor 1001, such as a CPU; a communication bus 1002; a user interface 1003; a network interface 1004; and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0040] Those skilled in the art will understand that Figure 1 The device structure shown does not constitute a limitation on the device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0041] like Figure 1As shown, the memory 1005, which serves as a storage medium, may include an operating device, a network communication module, a user interface module, and a multi-machine dog collaborative task allocation program.

[0042] The device of the present invention calls the multi-machine dog cooperative task allocation program stored in the memory 1005 through the processor 1001, and performs the following operations: Based on the current time-varying communication topology, a particle neighborhood is constructed between each robot dog and a topology guidance center is calculated. The inertial weights are dynamically adjusted based on the particle neighborhood and the topology guidance center. The particle velocity and position are updated based on the adjusted inertia weights, and a hierarchical constraint process is applied to the updated particle positions. Based on the particle positions after constraint processing, a dynamic weight aggregation mechanism is used to balance the multi-objective function, and the environmental state vector is monitored in real time to trigger adaptive parameter adjustment. At the same time, when convergence stalls, a hybrid local search strategy is activated to control each robot dog to search.

[0043] The device of the present invention, through processor 1001 calling the multi-machine dog cooperative task allocation program stored in memory 1005, also performs the following operations: Obtain the current time-varying communication topology map, and construct a particle neighborhood for each robot dog based on the current time-varying communication topology map; The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score. The inertia weight is dynamically adjusted based on the size of the particle neighborhood and the distribution characteristics of the topological guidance center.

[0044] The device of the present invention, through processor 1001 calling the multi-machine dog cooperative task allocation program stored in memory 1005, also performs the following operations: Obtain the vertex set, communication link edge set, and edge weight communication quality score of the robot dog, and determine the current time-varying communication topology using the following formula:

[0045] in, This is the current time-varying communication topology diagram. For the vertex set of the robot dog, For communication link edge set, Scoring the quality of edge-weighted communication; Real-time collection of status data for each robot dog; dynamic updating of the topology map based on the status data. To select neighboring robot dogs that meet the communication quality score from the set of robot dog vertices for each robot dog, the particle neighborhood is constructed using the following formula:

[0046] in, For the particle neighborhood, For the neighbor's robot dog, For the vertex set of the robot dog, For robot dogs and Inter-link communication quality score, This is a preset scoring threshold.

[0047] The device of the present invention, through processor 1001 calling the multi-machine dog cooperative task allocation program stored in memory 1005, also performs the following operations: The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score, according to the following formula:

[0048] in, As the topology guidance center, For the particle neighborhood, For robot dogs and Inter-link communication quality score, For neighborhood particles Its historical best position.

[0049] The device of the present invention, through processor 1001 calling the multi-machine dog cooperative task allocation program stored in memory 1005, also performs the following operations: The particle velocity is updated according to the adjusted inertia weight using the following formula:

[0050] in, For particles In the The speed of generation For particles In the The inertial weight of the generation, For particles In the The speed of generation Learning factors that control for the influence of individual experience. A random number in the interval [0,1]. For particles The best historical position For particles In the The position of the generation, To control the learning factor of neighborhood guidance influence, A random number in the interval [0,1]. As the topology guidance center, To constrain the penalty coefficient, To constrain the penalties for violations; Perform a position update and obtain the updated particle position using the following formula:

[0051] in, For particles In the The updated position of the generation For particles In the The position of the generation, For particles In the The speed of generation; The feasible region is mapped by hard constraint projection using the following formula:

[0052] in, For particles In the The updated position of the generation For projection operators, This is the feasible region with hard constraints; The penalty term for the degree of violation of soft constraints is applied by the following formula:

[0053] in, To constrain the penalties for violations, For the first The penalty coefficient for a soft constraint, For particles For the The degree of violation of a soft constraint; Heuristic repair strategies are used to reassign tasks, reallocate resources, fine-tune timing, and switch emergency modes for particles that violate constraints.

[0054] The device of the present invention, through processor 1001 calling the multi-machine dog cooperative task allocation program stored in memory 1005, also performs the following operations: Based on the particle positions after constraint processing, the weights of each sub-target are adaptively calculated using the following formula through a dynamic weight aggregation mechanism:

[0055] in, For the first The sub-goals in the The weight of generations For the first The dynamic importance of individual sub-goals; The environment state vector is obtained using the following formula:

[0056] in, For the first The environment state vector of the generation, For changes in network topology, For changes in the task set, This refers to the changes in the robot dog's state. The dynamic importance of each sub-objective is dynamically adjusted based on the environmental state vector and historical optimization performance, thereby balancing the multi-objective function in the following equation:

[0057]

[0058]

[0059]

[0060]

[0061] in, For particles The comprehensive objective function value, The dynamic weights of the sub-objective function for the total time. Let be a sub-objective function of the total time. The dynamic weights of the sub-objective function for total energy consumption. Let be a sub-objective function of total energy consumption. The dynamic weights are the sub-objective functions for load balancing. This is a sub-objective function for load balancing. The dynamic weights of the sub-objective function for time limit violations. For the sub-objective function of time limit violation, To allocate matrix elements, For robot dogs Execute the task Required completion time For robot dogs Execute the task energy consumption For the task For robot dogs Workload weighting It is the variance function. For the task The actual completion time, For the task The deadline This indicates that if the task expires, the expiration time will be returned; otherwise, 0 will be returned. The system monitors the environmental state vector in real time. When the environmental change detection condition is triggered, it dynamically adjusts each parameter and continuously evaluates the convergence state of the algorithm. When convergence stalls, it activates a hybrid local search strategy to control each robot dog to search.

[0062] The device of the present invention, through processor 1001 calling the multi-machine dog cooperative task allocation program stored in memory 1005, also performs the following operations: Real-time monitoring of the environmental state vector, and determination of environmental change detection conditions using the following formula:

[0063] in, Let be the environment state vector at the current time t. Let be the environment state vector at the previous time t-1. For dynamic adaptive thresholds; When the environmental mutation detection condition is triggered, the inertia weight, the learning factor that controls the influence of individual experience, the learning factor that controls the influence of neighborhood guidance, and the particle neighborhood size are dynamically adjusted. The algorithm's convergence status is continuously evaluated. When the inverse generational distance or hypervolume index of the Pareto optimal solution set improves below a preset threshold for N consecutive generations, a hybrid local search strategy is adaptively activated. Each robot dog is controlled to search according to the hybrid local search strategy, which includes greedy improvement, simulated annealing, variable neighborhood search, or tabu search.

[0064] This embodiment constructs particle neighborhoods among the robot dogs using the current time-varying communication topology graph and calculates the topology guidance center. Inertial weights are dynamically adjusted based on these neighborhoods and guidance centers. Particle velocities and positions are updated according to the adjusted inertial weights, and hierarchical constraint processing is applied to the updated particle positions. Based on the constrained particle positions, a dynamic weight aggregation mechanism is used to balance multiple objective functions, and environmental state vectors are monitored in real time to trigger adaptive parameter adjustments. Simultaneously, a hybrid local search strategy is activated when convergence stalls to control each robot dog's search. This ensures the feasibility of the task allocation scheme, significantly improving the efficiency, feasibility, and environmental adaptability of multi-robot collaborative task allocation. It enhances the optimization quality of the task allocation scheme, strengthens the algorithm's adaptability to dynamic network topology and communication quality changes, ensures the allocation scheme meets the hard constraints of kinematics, energy, and communication of the physical robot, improves the algorithm's convergence speed and solution set diversity, avoids getting trapped in local optima, adapts to complex dynamic environments, ensures the robustness and real-time performance of the allocation scheme, and improves the speed and efficiency of multi-robot collaborative task allocation.

[0065] Based on the above hardware structure, an embodiment of the multi-robot collaborative task allocation method of the present invention is proposed.

[0066] Reference Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the multi-robot collaborative task allocation method of the present invention.

[0067] In the first embodiment, the multi-robot collaborative task allocation method includes the following steps: Step S10: Construct particle neighborhoods between each robot dog based on the current time-varying communication topology graph and calculate the topology guidance center. Dynamically adjust the inertia weights based on the particle neighborhoods and the topology guidance center.

[0068] It should be noted that, based on the current time-varying communication topology, the particle neighborhood of each robot dog can be dynamically constructed, and the topology guidance center can be calculated. The inertia weight can be dynamically adjusted according to the particle neighborhood and the topology guidance center.

[0069] Step S20: Update the particle velocity and particle position according to the adjusted inertia weight, and perform hierarchical constraint processing on the updated particle position.

[0070] It should be understood that the particle velocity and position updates are calculated based on dynamically adjusted inertial weights, and hierarchical constraint processing is performed on the new positions, thereby generating a feasible and efficient multi-robot dog cooperative task allocation scheme.

[0071] Step S30: Based on the particle positions after constraint processing, a dynamic weight aggregation mechanism is used to balance the multi-objective function, and the environmental state vector is monitored in real time to trigger adaptive adjustment of parameters. At the same time, when convergence stalls, a hybrid local search strategy is activated to control each robot dog to search.

[0072] Understandably, based on the particle positions after constraint processing, a dynamic weight aggregation mechanism is used to adaptively balance multiple objective functions such as task completion rate, energy consumption, and latency. The environmental state vector is monitored in real time to trigger adaptive adjustments of various parameters. When the algorithm converges and stagnates, a hybrid local search strategy is activated to control each robot dog to search, which significantly improves the global efficiency and environmental adaptability of multi-robot collaborative task allocation.

[0073] This embodiment constructs particle neighborhoods among the robot dogs using the current time-varying communication topology graph and calculates the topology guidance center. Inertial weights are dynamically adjusted based on these neighborhoods and guidance centers. Particle velocities and positions are updated according to the adjusted inertial weights, and hierarchical constraint processing is applied to the updated particle positions. Based on the constrained particle positions, a dynamic weight aggregation mechanism is used to balance multiple objective functions, and environmental state vectors are monitored in real time to trigger adaptive parameter adjustments. Simultaneously, a hybrid local search strategy is activated when convergence stalls to control each robot dog's search. This ensures the feasibility of the task allocation scheme, significantly improving the efficiency, feasibility, and environmental adaptability of multi-robot collaborative task allocation. It enhances the optimization quality of the task allocation scheme, strengthens the algorithm's adaptability to dynamic network topology and communication quality changes, ensures the allocation scheme meets the hard constraints of kinematics, energy, and communication of the physical robot, improves the algorithm's convergence speed and solution set diversity, avoids getting trapped in local optima, adapts to complex dynamic environments, ensures the robustness and real-time performance of the allocation scheme, and improves the speed and efficiency of multi-robot collaborative task allocation.

[0074] Furthermore, Figure 3 This is a flowchart illustrating the second embodiment of the multi-robot collaborative task allocation method of the present invention, as shown below. Figure 3 As shown, based on the first embodiment, a second embodiment of the multi-robot collaborative task allocation method of the present invention is proposed. In this embodiment, step S10 specifically includes the following steps: Step S11: Obtain the current time-varying communication topology map, and construct a particle neighborhood for each robot dog based on the current time-varying communication topology map.

[0075] It should be noted that, based on the real-time acquired time-varying communication topology map, neighboring robot dogs whose communication quality scores exceed a preset threshold are selected, and each robot dog dynamically constructs a particle neighborhood to reflect its local connectivity status in the current communication environment.

[0076] Furthermore, step S11 specifically includes the following steps: Obtain the vertex set, communication link edge set, and edge weight communication quality score of the robot dog, and determine the current time-varying communication topology using the following formula:

[0077] in, This is the current time-varying communication topology diagram. For the vertex set of the robot dog, For communication link edge set, Scoring the quality of edge-weighted communication; Real-time collection of status data for each robot dog; dynamic updating of the topology map based on the status data. To select neighboring robot dogs that meet the communication quality score from the set of robot dog vertices for each robot dog, the particle neighborhood is constructed using the following formula:

[0078] in, For the particle neighborhood, For the neighbor's robot dog, For the vertex set of the robot dog, For robot dogs and Inter-link communication quality score, This is a preset scoring threshold.

[0079] Understandably, this is based on the current time-varying communication topology. The system collects real-time status data such as signal strength and delay of each robot dog via wireless communication protocols, and dynamically updates the topology map; subsequently, it performs data collection for each robot dog. From vertex set Select those that meet the communication quality score ( Neighbor robot dog (with a preset scoring threshold) Constructing particle neighborhoods This neighborhood dynamically reflects the local connectivity state of the robot dog in a time-varying communication environment, providing a foundation for subsequent topology-aware optimization. This represents the set of actual communication links between the robot dogs (e.g., ), This indicates the quality score of the communication link.

[0080] Step S12: Calculate the topology guidance center using the historical best position of particles in the particle's neighborhood and the communication quality score.

[0081] It should be understood that, based on the historical best position and communication quality score of particles in the particle's neighborhood, a topological guidance center can be calculated using a weighted average formula, so that the guidance center focuses on neighbors with high communication quality, thereby dynamically optimizing the global search direction of the particle swarm in a time-varying communication environment.

[0082] The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score, according to the following formula:

[0083] in, As the topology guidance center, For the particle neighborhood, For robot dogs and Inter-link communication quality score, For neighborhood particles Its historical best position.

[0084] It should be noted that, based on particle neighborhood The historical best positions of each particle and communication quality score The topology guidance center is calculated using the weighted average formula described above, where the communication quality score is... As a weighting coefficient, the topology guidance center prioritizes focusing on neighboring particles with high communication quality, thereby dynamically optimizing the inertia weight. To enhance the diversity and global search capability of particle swarms in time-varying communication topologies.

[0085] Step S13: Dynamically adjust the inertia weight according to the size of the particle neighborhood and the distribution characteristics of the topological guidance center.

[0086] Understandably, based on the size of the particle neighborhood and the distribution characteristics of the topological guidance center, the inertial weight can be dynamically adjusted to enhance the global search diversity and convergence efficiency of the particle swarm in a time-varying communication environment in real time.

[0087] Furthermore, step S20 specifically includes the following steps: The particle velocity is updated according to the adjusted inertia weight using the following formula:

[0088] in, For particles In the The speed of generation For particles In the The inertial weight of the generation, For particles In the The speed of generation Learning factors that control for the influence of individual experience. A random number in the interval [0,1]. For particles The best historical position For particles In the The position of the generation, To control the learning factor of neighborhood guidance influence, A random number in the interval [0,1]. As the topology guidance center, To constrain the penalty coefficient, To constrain the penalties for violations; Perform a position update and obtain the updated particle position using the following formula:

[0089] in, For particles In the The updated position of the generation For particles In the The position of the generation, For particles In the The speed of generation; The feasible region is mapped by hard constraint projection using the following formula:

[0090] in, For particles In the The updated position of the generation For projection operators, This is the feasible region with hard constraints; The penalty term for the degree of violation of soft constraints is applied by the following formula:

[0091] in, To constrain the penalties for violations, For the first The penalty coefficient for a soft constraint, For particles For the The degree of violation of a soft constraint; Heuristic repair strategies are used to reassign tasks, reallocate resources, fine-tune timing, and switch emergency modes for particles that violate constraints.

[0092] It should be noted that, and The learning factor controls for the influence of individual experience and neighborhood guidance, with a typical range of [1.5, 2.5]. and The numbers are random and uniformly distributed in [0, 1], enhancing random exploration. Calculated from the weighted positions of particles in the neighborhood. To constrain the penalty coefficient, adjust the impact of constraint violation on speed. Penalties for constraint violations are based on a comprehensive evaluation of both hard and soft constraints. This indicates the updated task allocation scheme. This will map the particle position to the hard-constrained feasible region. , To constrain the feasible domain, constraints are included, such as unique task allocation, kinematic upper bound, and communication connectivity.

[0093] It is understandable that the hierarchical constraint handling mechanism is divided into three levels: hard constraint projection, soft constraint penalty, and heuristic repair. The degree of violation includes exceeding energy consumption limits, slight delay exceeding limits, etc. This embodiment does not impose any restrictions on these. Heuristic repair: For particles that violate constraints, the following operator is applied: Task reassignment: Reassign infeasible tasks to robot dogs that meet the constraints.

[0094] Resource reallocation: Optimizing the allocation ratio of computing or energy resources.

[0095] Timing fine-tuning: Adjust the task execution order to meet the deadline.

[0096] Emergency mode switching: Switch to low power or low priority mode in extreme situations.

[0097] Furthermore, step S30 specifically includes the following steps: Based on the particle positions after constraint processing, the weights of each sub-target are adaptively calculated using the following formula through a dynamic weight aggregation mechanism:

[0098] in, For the first The sub-goals in the The weight of generations For the first The dynamic importance of individual sub-goals; The environment state vector is obtained using the following formula:

[0099] in, For the first The environment state vector of the generation, For changes in network topology, For changes in the task set, This refers to the changes in the robot dog's state. The dynamic importance of each sub-objective is dynamically adjusted based on the environmental state vector and historical optimization performance, thereby balancing the multi-objective function in the following equation:

[0100]

[0101]

[0102]

[0103]

[0104] in, For particles The comprehensive objective function value, The dynamic weights of the sub-objective function for the total time. Let be a sub-objective function of the total time. The dynamic weights of the sub-objective function for total energy consumption. Let be a sub-objective function of total energy consumption. The dynamic weights are the sub-objective functions for load balancing. This is a sub-objective function for load balancing. The dynamic weights of the sub-objective function for time limit violations. For the sub-objective function of time limit violation, To allocate matrix elements, For robot dogs Execute the task Required completion time For robot dogs Execute the task energy consumption For the task For robot dogs Workload weighting It is the variance function. For the task The actual completion time, For the task The deadline This indicates that if the task expires, the expiration time will be returned; otherwise, 0 will be returned. The system monitors the environmental state vector in real time. When the environmental change detection condition is triggered, it dynamically adjusts each parameter and continuously evaluates the convergence state of the algorithm. When convergence stalls, it activates a hybrid local search strategy to control each robot dog to search.

[0105] It should be noted that, in order to balance the conflicts among multiple objectives, the following strategies can be adopted: Dynamic weight aggregation is performed using the formula described above, where... Adaptive adjustment based on environmental changes and historical performance.

[0106] Pareto Archive: Maintain a non-dominated solution set A. When the archive overflows, similar solutions are eliminated based on the crowding distance in the target space and the diversity of solutions.

[0107] Multiple leader selection: Randomly sample from the Pareto archive based on crowding distance and quality score, and dynamically assign a leader to each particle. This enhances the ability to explore the entire system.

[0108] In practical implementation, to overcome the single-point learning and lack of diversity problems of traditional PSO, the neighborhood of particles can be dynamically constructed and the learning behavior can be adjusted in combination with communication quality.

[0109] For particles The current position and velocity distribution in generation t are denoted as . and The individual's historical best position is The neighborhood of the particle Based on the current communication topology and link quality Adaptive build.

[0110] Adaptive adjustment of inertia weight:

[0111] in, Let be the inertial weight of particle i in generation t, controlling the degree to which the particle retains its previous velocity. and The upper and lower bounds of the inertia weight are typically defined as follows: , , To account for the current population diversity, calculations are performed based on the dispersion of particle positions. To maximize group diversity, it is used for normalization. The adjustment factor controls the rate at which the weights decay with the iteration number t, where T is the maximum number of iterations.

[0112] It should be noted that the goal of task allocation is to comprehensively optimize the following four sub-objectives, forming a multi-objective optimization problem: To evaluate the overall merits of the allocation scheme, the dynamic weights of each sub-objective must satisfy the following:

[0113] Used to balance the importance of different objectives.

[0114] It can minimize the total latency of task completion. It can minimize the energy consumption of all robot dogs. It can minimize the variance of task allocation among the robot dogs. It can minimize the total task overdue. To assign matrix elements, representing tasks Should it be assigned to the robot dog? , For robot dogs Execute the task The required completion time depends on the task requirements. Robot dog capabilities And the distance traveled, For robot dogs Execute the task Energy consumption, taking into account mobile energy consumption, computing energy consumption, and communication energy consumption. For the task For robot dogs The workload weight reflects the complexity and resource requirements of the task. This is a variance function that measures the uniformity of the workload distribution among the robot dogs. For the task The actual completion time is determined by the allocation plan and the execution order.

[0115] Furthermore, the steps involve real-time monitoring of the environmental state vector. When a sudden environmental change detection condition is triggered, the parameters are dynamically adjusted, and the algorithm's convergence state is continuously evaluated. When convergence stalls, a hybrid local search strategy is activated to control each robot dog to perform the search, including: Real-time monitoring of the environmental state vector, and determination of environmental change detection conditions using the following formula:

[0116] in, Let be the environment state vector at the current time t. Let be the environment state vector at the previous time t-1. For dynamic adaptive thresholds; When the environmental mutation detection condition is triggered, the inertia weight, the learning factor that controls the influence of individual experience, the learning factor that controls the influence of neighborhood guidance, and the particle neighborhood size are dynamically adjusted. The algorithm's convergence status is continuously evaluated. When the inverse generational distance or hypervolume index of the Pareto optimal solution set improves below a preset threshold for N consecutive generations, a hybrid local search strategy is adaptively activated. Each robot dog is controlled to search according to the hybrid local search strategy, which includes greedy improvement, simulated annealing, variable neighborhood search, or tabu search.

[0117] It should be noted that, in order to cope with environmental changes (such as equipment failure and path congestion) in dynamic scenarios such as thermal power plant inspections, an environmental perception and adaptive adjustment mechanism is introduced: Environmental Vector:

[0118] in: : The environment state vector of generation t.

[0119] Changes in network topology (such as adding / disconnecting communication edges).

[0120] Changes to the task set (such as adding new tasks or adjusting priorities).

[0121] Changes in the robot dog's status (such as decreased battery level or location change).

[0122] Environmental mutation detection: If This triggers an "environmental mutation," where the threshold is... Determined adaptively based on historical statistics.

[0123] Adaptive adjustment: Update inertia weights and learning factors and ; Adjust the probability and size of neighboring edges to increase population diversity; If the environment changes drastically, some particles are randomly restarted to avoid getting trapped in local optima.

[0124] Understandably, to improve local mining capabilities, this embodiment introduces a hybrid local search strategy, including the following strategy pool: Greedy Improvement: Perform local optimization on the current solution, prioritizing the adjustment of high-cost tasks.

[0125] Simulated annealing: accepts suboptimal solutions with probability in order to escape local optima.

[0126] Variable neighborhood search: Dynamically adjusts the search range to balance exploration and utilization.

[0127] Taboo search: Records visited solutions to avoid duplicate searches.

[0128] The strategy selection is adaptively triggered based on the algorithm's stagnation time and convergence. If a local optimum trap is detected, a mutation operation is performed and the search step size is increased to escape the trap.

[0129] In the specific implementation, the algorithm flow (pseudocode) enter: Let D be the set of robot dogs, T be the set of tasks, G(t) be the time-varying topology, and C be the set of constraints.

[0130] Output: Pareto optimal task allocation solution set A.

[0131] 1. Initialize the particle swarm ,speed Individual historical best Pareto Archives Set the algorithm parameters.

[0132] 2. Loop (iteration t=1 to...) ): 3. Monitor environmental changes and update adaptive parameters (such as...) , and ) and neighborhood topology .

[0133] 4. For each particle i: 5. Evaluate multi-objective functions Update individual optimal And Pareto Archive A.

[0134] 6. End particle cycle.

[0135] 7. Select multiple leaders from file A and assign a topological guidance center to each particle. .

[0136] 8. For each particle i: 9. Calculate the topology sensing speed. Update location .

[0137] 10. Perform hierarchical constraint processing (projection, penalty, repair).

[0138] 11. End particle cycle.

[0139] 12. If a local search condition is triggered, an enhancement policy is selected from the policy pool and executed.

[0140] 13. If the convergence criterion is met (e.g., no significant improvement after multiple generations of Pareto front), then exit the loop.

[0141] 14. End the iteration.

[0142] 15. Return to the Pareto optimal solution set A.

[0143] To ensure algorithm performance, the following parameter settings are recommended: Population size: Adjustments will be made based on the scale of the problem.

[0144] Maximum number of iterations: .

[0145] Learning factor range: .

[0146] Inertia weight range: .

[0147] Pareto archive size: .

[0148] Neighborhood size: .

[0149] Local search trigger probability: .

[0150] Convergence criterion: No significant improvement in the Pareto front for 10–20 consecutive generations (e.g., the inverted generational distance (IGD) or hypervolume (HV) index is below the threshold).

[0151] It should be noted that the set of robot dogs is represented as: , where m is the total number of robot dogs.

[0152] Each robot dog It has the following attributes: Computational / Execution Capabilities This reflects the computational performance of the task it processes (such as CPU frequency and sensor accuracy). Location , representing the robot dog's current position coordinates in two-dimensional or three-dimensional space; Remaining battery power This represents the currently available energy, which affects the sustainability of task execution.

[0153] Task set: represented as , where n is the total number of tasks.

[0154] Each task have: Computational / Perception Requirements This represents the amount of computational or perceptual resources required to complete the task; Deadline That is, the deadline by which the task must be completed; Priority This is used to prioritize high-priority tasks in task allocation.

[0155] Network topology: using time-varying undirected graphs It means that, among them: Vertex set This applies to all robot dogs; Edge set This indicates the communication connection between the robot dogs; Border rights Represents a robot dog and Communication quality between them (such as a comprehensive score based on signal-to-noise ratio (SNR), throughput, or latency).

[0156] Allocation matrix: defined as ,in Indicates task Assigned to robot dog ,otherwise .

[0157] Particle encoding: Each particle A complete task allocation plan A is provided, which includes detailed information on execution order, path planning, and timing arrangements to ensure the operability of the allocation plan.

[0158] Basic feasibility constraints: Task uniqueness: Each task Only one robot dog is assigned, that is .

[0159] Capability Constraints: Robot Dog The total task requirements must not exceed its computing power. .

[0160] Energy Constraint: The energy consumption for task execution must not exceed .

[0161] Communication constraints: The allocation scheme must ensure communication connectivity between the robot dogs performing the tasks.

[0162] Compared with traditional Particle Swarm Optimization (PSO) and Multi-Objective Particle Swarm Optimization (MOPSO) algorithms, this embodiment demonstrates significant advantages in the following aspects: Solution set quality: Through multi-objective dynamic weights and Pareto archive mechanism, comprehensive optimization of total latency, energy consumption, load balancing and time limit violations is achieved.

[0163] Convergence speed: Topology-aware speed updates and hybrid local search significantly accelerate convergence while avoiding local optimum traps.

[0164] Understanding diversity: Multiple leader selection and dynamic neighborhood adjustment enhance the ability to comprehensively explore the understanding space.

[0165] Feasibility of constraints: Hierarchical constraint processing and heuristic repair ensure that the allocation scheme meets hard constraints (such as task uniqueness and communication connectivity) and soft constraints (such as energy consumption limits).

[0166] Dynamic adaptability: Environmental mutation detection and adaptive parameter adjustment enable the algorithm to cope with complex dynamic scenarios such as thermal power plant inspections.

[0167] This embodiment, through the above-described scheme, obtains the current time-varying communication topology map, constructs particle neighborhoods for each robot dog based on the current time-varying communication topology map, calculates the topology guidance center using the historical best position of particles within the particle neighborhood and the communication quality score, and dynamically adjusts the inertia weights according to the size of the particle neighborhood and the distribution characteristics of the topology guidance center. This significantly improves the global search diversity and convergence efficiency of the particle swarm under communication link fluctuations, effectively avoids the algorithm getting trapped in local optima, and provides a highly robust optimization foundation for multi-robot dog collaborative task allocation.

[0168] Accordingly, the present invention further provides a multi-robot dog collaborative task allocation device.

[0169] Reference Figure 4 , Figure 4 This is a functional block diagram of the first embodiment of the multi-robot collaborative task allocation device of the present invention.

[0170] In a first embodiment of the multi-robot collaborative task allocation device of the present invention, the multi-robot collaborative task allocation device includes: The dynamic adjustment module 10 is used to construct the particle neighborhood between each robot dog according to the current time-varying communication topology map and calculate the topology guidance center, and dynamically adjust the inertial weight according to the particle neighborhood and the topology guidance center.

[0171] The hierarchical constraint module 20 is used to update the particle velocity and particle position according to the adjusted inertia weight, and to perform hierarchical constraint processing on the updated particle position.

[0172] The control module 30 is adjusted to balance the multi-objective function using a dynamic weight aggregation mechanism based on the particle position after constraint processing, and to monitor the environmental state vector in real time to trigger parameter adaptive adjustment. At the same time, when convergence stalls, a hybrid local search strategy is activated to control each robot dog to search.

[0173] The dynamic adjustment module 10 is also used to acquire the current time-varying communication topology map, construct a particle neighborhood for each robot dog based on the current time-varying communication topology map, calculate the topology guidance center using the historical best position of the particles in the particle neighborhood and the communication quality score, and dynamically adjust the inertia weight based on the size of the particle neighborhood and the distribution characteristics of the topology guidance center.

[0174] The dynamic adjustment module 10 is also used to obtain the robot dog vertex set, communication link edge set, and edge weight communication quality score, and determine the current time-varying communication topology using the following formula:

[0175] in, This is the current time-varying communication topology diagram. For the vertex set of the robot dog, For communication link edge set, Scoring the quality of edge-weighted communication; Real-time collection of status data for each robot dog; dynamic updating of the topology map based on the status data. To select neighboring robot dogs that meet the communication quality score from the set of robot dog vertices for each robot dog, the particle neighborhood is constructed using the following formula:

[0176] in, For the particle neighborhood, For the neighbor's robot dog, For the vertex set of the robot dog, For robot dogs and Inter-link communication quality score, This is a preset scoring threshold.

[0177] The dynamic adjustment module 10 is also used to calculate the topology guidance center using the historical best position and communication quality score of particles in the particle's neighborhood through the following formula:

[0178] in, As the topology guidance center, For the particle neighborhood, For robot dogs and Inter-link communication quality score, For neighborhood particles Its historical best position.

[0179] The hierarchical constraint module 20 is also used to update the particle velocity according to the adjusted inertia weight using the following formula:

[0180] in, For particles In the The speed of generation For particles In the The inertial weight of the generation, For particles In the The speed of generation Learning factors that control for the influence of individual experience. A random number in the interval [0,1]. For particles The best historical position For particles In the The position of the generation, To control the learning factor of neighborhood guidance influence, A random number in the interval [0,1]. As the topology guidance center, To constrain the penalty coefficient, To constrain the penalties for violations; Perform a position update and obtain the updated particle position using the following formula:

[0181] in, For particles In the The updated position of the generation For particles In the The position of the generation, For particles In the The speed of generation; The feasible region is mapped by hard constraint projection using the following formula:

[0182] in, For particles In the The updated position of the generation For projection operators, This is the feasible region with hard constraints; The penalty term for the degree of violation of soft constraints is applied by the following formula:

[0183] in, To constrain the penalties for violations, For the first The penalty coefficient for a soft constraint, For particles For the The degree of violation of a soft constraint; Heuristic repair strategies are used to reassign tasks, reallocate resources, fine-tune timing, and switch emergency modes for particles that violate constraints.

[0184] The adjustment control module 30 is also used to adaptively calculate the weights of each sub-target based on the particle positions after constraint processing, using a dynamic weight aggregation mechanism and the following formula:

[0185] in, For the first The sub-goals in the The weight of generations For the first The dynamic importance of individual sub-goals; The environment state vector is obtained using the following formula:

[0186] in, For the first The environment state vector of the generation, For changes in network topology, For changes in the task set, This refers to the changes in the robot dog's state. The dynamic importance of each sub-objective is dynamically adjusted based on the environmental state vector and historical optimization performance, thereby balancing the multi-objective function in the following equation:

[0187]

[0188]

[0189]

[0190]

[0191] in, For particles The comprehensive objective function value, The dynamic weights of the sub-objective function for the total time. Let be a sub-objective function of the total time. The dynamic weights of the sub-objective function for total energy consumption. Let be a sub-objective function of total energy consumption. The dynamic weights are the sub-objective functions for load balancing. This is a sub-objective function for load balancing. The dynamic weights of the sub-objective function for time limit violations. For the sub-objective function of time limit violation, To allocate matrix elements, For robot dogs Execute the task Required completion time For robot dogs Execute the task energy consumption For the task For robot dogs Workload weighting It is the variance function. For the task The actual completion time, For the task The deadline This indicates that if the task expires, the expiration time will be returned; otherwise, 0 will be returned. The system monitors the environmental state vector in real time. When the environmental change detection condition is triggered, it dynamically adjusts each parameter and continuously evaluates the convergence state of the algorithm. When convergence stalls, it activates a hybrid local search strategy to control each robot dog to search.

[0192] The adjustment control module 30 is also used to monitor the environmental state vector in real time and determine the environmental change detection conditions using the following formula:

[0193] in, Let be the environment state vector at the current time t. Let be the environment state vector at the previous time t-1. For dynamic adaptive thresholds; When the environmental mutation detection condition is triggered, the inertia weight, the learning factor that controls the influence of individual experience, the learning factor that controls the influence of neighborhood guidance, and the particle neighborhood size are dynamically adjusted. The algorithm's convergence status is continuously evaluated. When the inverse generational distance or hypervolume index of the Pareto optimal solution set improves below a preset threshold for N consecutive generations, a hybrid local search strategy is adaptively activated. Each robot dog is controlled to search according to the hybrid local search strategy, which includes greedy improvement, simulated annealing, variable neighborhood search, or tabu search.

[0194] The steps for implementing each functional module of the multi-robot collaborative task allocation device can be referred to in the various embodiments of the multi-robot collaborative task allocation method of the present invention, and will not be repeated here.

[0195] Furthermore, embodiments of the present invention also propose a storage medium storing a multi-machine dog collaborative task allocation program, which, when executed by a processor, performs the following operations: Based on the current time-varying communication topology, a particle neighborhood is constructed between each robot dog and a topology guidance center is calculated. The inertial weights are dynamically adjusted based on the particle neighborhood and the topology guidance center. The particle velocity and position are updated based on the adjusted inertia weights, and a hierarchical constraint process is applied to the updated particle positions. Based on the particle positions after constraint processing, a dynamic weight aggregation mechanism is used to balance the multi-objective function, and the environmental state vector is monitored in real time to trigger adaptive parameter adjustment. At the same time, when convergence stalls, a hybrid local search strategy is activated to control each robot dog to search.

[0196] Furthermore, when the multi-machine dog collaborative task allocation program is executed by the processor, it also performs the following operations: Obtain the current time-varying communication topology map, and construct a particle neighborhood for each robot dog based on the current time-varying communication topology map; The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score. The inertia weight is dynamically adjusted based on the size of the particle neighborhood and the distribution characteristics of the topological guidance center.

[0197] Furthermore, when the multi-machine dog collaborative task allocation program is executed by the processor, it also performs the following operations: Obtain the vertex set, communication link edge set, and edge weight communication quality score of the robot dog, and determine the current time-varying communication topology using the following formula:

[0198] in, This is the current time-varying communication topology diagram. For the vertex set of the robot dog, For communication link edge set, Scoring the quality of edge-weighted communication; Real-time collection of status data for each robot dog; dynamic updating of the topology map based on the status data. To select neighboring robot dogs that meet the communication quality score from the set of robot dog vertices for each robot dog, the particle neighborhood is constructed using the following formula:

[0199] in, For the particle neighborhood, For the neighbor's robot dog, For the vertex set of the robot dog, For robot dogs and Inter-link communication quality score, This is a preset scoring threshold.

[0200] Furthermore, when the multi-machine dog collaborative task allocation program is executed by the processor, it also performs the following operations: The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score, according to the following formula:

[0201] in, As the topology guidance center, For the particle neighborhood, For robot dogs and Inter-link communication quality score, For neighborhood particles Its historical best position.

[0202] Furthermore, when the multi-machine dog collaborative task allocation program is executed by the processor, it also performs the following operations: The particle velocity is updated according to the adjusted inertia weight using the following formula:

[0203] in, For particles In the The speed of generation For particles In the The inertial weight of the generation, For particles In the The speed of generation Learning factors that control for the influence of individual experience. A random number in the interval [0,1]. For particles The best historical position For particles In the The position of the generation, To control the learning factor of neighborhood guidance influence, A random number in the interval [0,1]. As the topology guidance center, To constrain the penalty coefficient, To constrain the penalties for violations; Perform a position update and obtain the updated particle position using the following formula:

[0204] in, For particles In the The updated position of the generation For particles In the The position of the generation, For particles In the The speed of generation; The feasible region is mapped by hard constraint projection using the following formula:

[0205] in, For particles In the The updated position of the generation For projection operators, This is the feasible region with hard constraints; The penalty term for the degree of violation of soft constraints is applied by the following formula:

[0206] in, To constrain the penalties for violations, For the first The penalty coefficient for a soft constraint, For particles For the The degree of violation of a soft constraint; Heuristic repair strategies are used to reassign tasks, reallocate resources, fine-tune timing, and switch emergency modes for particles that violate constraints.

[0207] Furthermore, when the multi-machine dog collaborative task allocation program is executed by the processor, it also performs the following operations: Based on the particle positions after constraint processing, the weights of each sub-target are adaptively calculated using the following formula through a dynamic weight aggregation mechanism:

[0208] in, For the first The sub-goals in the The weight of generations For the first The dynamic importance of individual sub-goals; The environment state vector is obtained using the following formula:

[0209] in, For the first The environment state vector of the generation, For changes in network topology, For changes in the task set, This refers to the changes in the robot dog's state. The dynamic importance of each sub-objective is dynamically adjusted based on the environmental state vector and historical optimization performance, thereby balancing the multi-objective function in the following equation:

[0210]

[0211]

[0212]

[0213]

[0214] in, For particles The comprehensive objective function value, The dynamic weights of the sub-objective function for the total time. Let be a sub-objective function of the total time. The dynamic weights of the sub-objective function for total energy consumption. Let be a sub-objective function of total energy consumption. The dynamic weights are the sub-objective functions for load balancing. This is a sub-objective function for load balancing. The dynamic weights of the sub-objective function for time limit violations. For the sub-objective function of time limit violation, To allocate matrix elements, For robot dogs Execute the task Required completion time For robot dogs Execute the task energy consumption For the task For robot dogs Workload weighting It is the variance function. For the task The actual completion time, For the task The deadline This indicates that if the task expires, the expiration time will be returned; otherwise, 0 will be returned. The system monitors the environmental state vector in real time. When the environmental change detection condition is triggered, it dynamically adjusts each parameter and continuously evaluates the convergence state of the algorithm. When convergence stalls, it activates a hybrid local search strategy to control each robot dog to search.

[0215] Furthermore, when the multi-machine dog collaborative task allocation program is executed by the processor, it also performs the following operations: Real-time monitoring of the environmental state vector, and determination of environmental change detection conditions using the following formula:

[0216] in, Let be the environment state vector at the current time t. Let be the environment state vector at the previous time t-1. For dynamic adaptive thresholds; When the environmental mutation detection condition is triggered, the inertia weight, the learning factor that controls the influence of individual experience, the learning factor that controls the influence of neighborhood guidance, and the particle neighborhood size are dynamically adjusted. The algorithm's convergence status is continuously evaluated. When the inverse generational distance or hypervolume index of the Pareto optimal solution set improves below a preset threshold for N consecutive generations, a hybrid local search strategy is adaptively activated. Each robot dog is controlled to search according to the hybrid local search strategy, which includes greedy improvement, simulated annealing, variable neighborhood search, or tabu search.

[0217] Those skilled in the art will understand that all or part of the steps in the methods described above can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium is a computer-readable storage medium, including: a USB flash drive, a portable hard drive, and a read-only memory (ROM). Various media that can store program code, such as only memory, random access memory (RAM), magnetic disks or optical disks.

[0218] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0219] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0220] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for task allocation in multi-robot collaboration, characterized in that, The multi-robot collaborative task allocation method includes: Based on the current time-varying communication topology, a particle neighborhood is constructed between each robot dog and a topology guidance center is calculated. The inertial weights are dynamically adjusted based on the particle neighborhood and the topology guidance center. The particle velocity and position are updated based on the adjusted inertia weights, and a hierarchical constraint process is applied to the updated particle positions. Based on the particle positions after constraint processing, a dynamic weight aggregation mechanism is used to balance the multi-objective function, and the environmental state vector is monitored in real time to trigger adaptive parameter adjustment. At the same time, when convergence stalls, a hybrid local search strategy is activated to control each robot dog to search.

2. The multi-robot collaborative task allocation method as described in claim 1, characterized in that, The step of constructing particle neighborhoods among each robot dog based on the current time-varying communication topology and calculating the topology guidance center, and dynamically adjusting the inertial weights based on the particle neighborhoods and the topology guidance center, includes: Obtain the current time-varying communication topology map, and construct a particle neighborhood for each robot dog based on the current time-varying communication topology map; The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score. The inertia weight is dynamically adjusted based on the size of the particle neighborhood and the distribution characteristics of the topological guidance center.

3. The multi-robot collaborative task allocation method as described in claim 2, characterized in that, The step of obtaining the current time-varying communication topology map and constructing particle neighborhoods for each robot dog based on the current time-varying communication topology map includes: Obtain the vertex set, communication link edge set, and edge weight communication quality score of the robot dog, and determine the current time-varying communication topology using the following formula: in, This is the current time-varying communication topology diagram. For the vertex set of the robot dog, For communication link edge set, Scoring the quality of edge-weighted communication; Real-time collection of status data for each robot dog; dynamic updating of the topology map based on the status data. To select neighboring robot dogs that meet the communication quality score from the set of robot dog vertices for each robot dog, the particle neighborhood is constructed using the following formula: in, For the particle neighborhood, For the neighbor's robot dog, For the vertex set of the robot dog, For robot dogs and Inter-link communication quality score, This is a preset scoring threshold.

4. The multi-robot collaborative task allocation method as described in claim 2, characterized in that, The calculation of the topology guidance center using the historical best position of particles in the particle's neighborhood and communication quality score includes: The topology guidance center is calculated using the historical best position of particles in the particle's neighborhood and the communication quality score, according to the following formula: in, As the topology guidance center, For the particle neighborhood, For robot dogs and Inter-link communication quality score, For neighborhood particles Its historical best position.

5. The multi-robot collaborative task allocation method as described in claim 1, characterized in that, The step of updating particle velocity and position based on the adjusted inertia weight, and performing hierarchical constraint processing on the updated particle position, includes: The particle velocity is updated according to the adjusted inertia weight using the following formula: in, For particles In the The speed of generation For particles In the The inertial weight of the generation, For particles In the The speed of generation Learning factors that control for the influence of individual experience. A random number in the interval [0,1]. For particles The best historical position For particles In the The position of the generation, To control the learning factor of neighborhood guidance influence, A random number in the interval [0,1]. As the topology guidance center, To constrain the penalty coefficient, To constrain the penalties for violations; Perform a position update and obtain the updated particle position using the following formula: in, For particles In the The updated position of the generation For particles In the The position of the generation, For particles In the The speed of generation; The feasible region is mapped by hard constraint projection using the following formula: in, For particles In the The updated position of the generation For projection operators, This is the feasible region with hard constraints; The penalty term for the degree of violation of soft constraints is applied by the following formula: in, To constrain the penalties for violations, For the first The penalty coefficient for a soft constraint, For particles For the The degree of violation of a soft constraint; Heuristic repair strategies are used to reassign tasks, reallocate resources, fine-tune timing, and switch emergency modes for particles that violate constraints.

6. The multi-robot collaborative task allocation method as described in claim 1, characterized in that, The particle positions, after constraint processing, utilize a dynamic weight aggregation mechanism to balance multi-objective functions and monitor the environmental state vector in real time to trigger adaptive parameter adjustments. Simultaneously, when convergence stalls, a hybrid local search strategy is activated to control each robot dog's search, including: Based on the particle positions after constraint processing, the weights of each sub-target are adaptively calculated using the following formula through a dynamic weight aggregation mechanism: in, For the first The sub-goals in the The weight of generations For the first The dynamic importance of individual sub-goals; The environment state vector is obtained using the following formula: in, For the first The environment state vector of the generation, For changes in network topology, For changes in the task set, This refers to the changes in the robot dog's state. The dynamic importance of each sub-objective is dynamically adjusted based on the environmental state vector and historical optimization performance, thereby balancing the multi-objective function in the following equation: in, For particles The comprehensive objective function value, The dynamic weights of the sub-objective function for the total time. Let be a sub-objective function of the total time. The dynamic weights of the sub-objective function for total energy consumption are... Let be a sub-objective function of total energy consumption. The dynamic weights are the sub-objective functions for load balancing. This is a sub-objective function for load balancing. The dynamic weights of the sub-objective function for time limit violations. For the sub-objective function of time limit violation, To allocate matrix elements, For robot dogs Execute the task Required completion time For robot dogs Execute the task energy consumption For the task For robot dogs Workload weighting It is the variance function. For the task The actual completion time For the task The deadline This indicates that if the task expires, the expiration time will be returned; otherwise, 0 will be returned. The system monitors the environmental state vector in real time. When the environmental change detection condition is triggered, it dynamically adjusts each parameter and continuously evaluates the convergence state of the algorithm. When convergence stalls, it activates a hybrid local search strategy to control each robot dog to search.

7. The multi-robot collaborative task allocation method as described in claim 6, characterized in that, The real-time monitoring environment state vector dynamically adjusts various parameters when the environment change detection condition is triggered, continuously evaluates the algorithm's convergence state, and activates a hybrid local search strategy to control each robot dog to search when convergence stalls, including: Real-time monitoring of the environmental state vector, and determination of environmental change detection conditions using the following formula: in, Let be the environment state vector at the current time t. Let be the environment state vector at the previous time t-1. For dynamic adaptive thresholds; When the environmental mutation detection condition is triggered, the inertia weight, the learning factor that controls the influence of individual experience, the learning factor that controls the influence of neighborhood guidance, and the particle neighborhood size are dynamically adjusted. The algorithm's convergence status is continuously evaluated. When the inverse generational distance or hypervolume index of the Pareto optimal solution set improves below a preset threshold for N consecutive generations, a hybrid local search strategy is adaptively activated. Each robot dog is controlled to search according to the hybrid local search strategy, which includes greedy improvement, simulated annealing, variable neighborhood search, or tabu search.

8. A multi-robot collaborative task allocation device, characterized in that, The multi-robot collaborative task allocation device includes: The dynamic adjustment module is used to construct particle neighborhoods between each robot dog based on the current time-varying communication topology map and calculate the topology guidance center, and dynamically adjust the inertial weights based on the particle neighborhoods and the topology guidance center; The hierarchical constraint module is used to update the particle velocity and particle position according to the adjusted inertia weight, and to perform hierarchical constraint processing on the updated particle position; The control module is adjusted to balance the multi-objective function using a dynamic weight aggregation mechanism based on the particle positions after constraint processing, and to monitor the environmental state vector in real time to trigger adaptive parameter adjustment. At the same time, when convergence stalls, a hybrid local search strategy is activated to control each robot dog to search.

9. A multi-robot collaborative task allocation device, characterized in that, The multi-robot collaborative task allocation device includes: a memory, a processor, and a multi-robot collaborative task allocation program stored in the memory and executable on the processor, wherein the multi-robot collaborative task allocation program is configured to implement the steps of the multi-robot collaborative task allocation method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a multi-robot collaborative task allocation program, which, when executed by a processor, implements the steps of the multi-robot collaborative task allocation method as described in any one of claims 1 to 7.

Images