Heterogeneous unmanned cluster cooperative control system based on edge computing

Abstract

The application discloses a heterogeneous unmanned cluster cooperative control system based on edge computing and belongs to the technical field of unmanned cluster cooperation, which comprises a management and control module, a situation awareness module, a joint optimization module and a decision feedback module. The management and control module is used for receiving the tasks issued by a user and generating initial scheduling instructions. The situation awareness module is used for acquiring the state data and environmental obstacle data of each unmanned device, combining the bearing data of an edge node and the unmanned device list under the jurisdiction, and constructing a regional situation set. The joint optimization module is used for constructing a joint optimization model, distributing and splitting the global joint optimization, solving sub-problems through iteration, adjusting the flight trajectory and task allocation plan of each unmanned device, combining the conflict checking and regulation and control of the edge node, and performing cooperative control on each unmanned device. The decision feedback module is used for continuously monitoring the environmental obstacle data and state data of each unmanned device and performing local re-planning.

Description

Technical Field

[0001] This invention relates to a heterogeneous unmanned swarm collaborative control system based on edge computing, belonging to the field of unmanned swarm collaborative technology. Background Technology

[0002] With the rapid development and widespread application of unmanned technology, single unmanned operation vehicles can no longer meet the increasingly complex task requirements. Heterogeneous unmanned swarms composed of various types of unmanned operation vehicles are gradually becoming the core development direction in fields such as emergency rescue and smart cities due to their significant advantages in coverage, functional complementarity and task flexibility.

[0003] A patent application with publication number CN116233790A discloses a 5G-based unmanned equipment cluster edge computing architecture system and processing method, relating to the field of communication technology. One specific implementation of the system includes: a command center, comprising a 5G core network and a transmission network; an airborne base station, comprising a 5G base station and an edge computing platform; and unmanned equipment, comprising a 5G terminal and sensors. The 5G core network and transmission network are connected to the 5G base station wirelessly. Sensors are used to collect service data, the 5G terminal is used to send the service data to its paired 5G base station, the 5G base station is used to send the service data to the edge computing platform, the edge computing platform is used to process the low-latency service data to form processed data and control commands; the 5G base station is used to send control commands to the 5G terminal, and the unmanned equipment is used to execute corresponding actions according to the control commands.

[0004] While existing technologies have improved the response speed and operational capabilities of future unmanned equipment clusters, they have not considered the strong coupling relationship between the flight trajectory and obstacle avoidance path of the unmanned equipment and the computational offloading tasks it undertakes. In actual three-dimensional space, changes in the flight path directly affect the quality of the communication link and the access selection of edge nodes, which in turn affects the loading efficiency and processing latency of the computational tasks. Conversely, the load characteristics and priorities of the computational tasks to be processed should also affect the path planning strategy of the unmanned equipment, such as whether it needs to be close to a specific edge node to meet low latency constraints. However, existing research often treats path planning and task offloading as two independent problems for separate optimization, failing to achieve joint decision-making and collaborative design between the two, resulting in suboptimal overall system performance. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a heterogeneous unmanned swarm collaborative control system based on edge computing. By constructing a regional situation set, configuring decision variables, and building a joint optimization model, it achieves collaborative decision-making on unmanned equipment flight trajectory and task allocation. Combined with edge node conflict control and local replanning, it breaks the limitations of independent optimization between the two and solves the technical problem of suboptimal system performance.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A heterogeneous unmanned swarm collaborative control system based on edge computing includes: a control module, a situational awareness module, a joint optimization module, and a decision feedback module;

[0008] The control module is used to receive tasks issued by users, obtain corresponding task data, and through preliminary matching and planning of unmanned equipment, edge nodes, and travel paths, combined with the travel paths to be taken by unmanned equipment, match the appropriate equipment type and associated nodes for the issued tasks, and generate initial scheduling instructions.

[0009] The situational awareness module responds to the initial scheduling command, acquires status data and environmental obstacle data, and combines the carrying data of the edge nodes and the list of unmanned devices under its jurisdiction to construct a regional situational awareness set for the edge nodes.

[0010] The joint optimization module is used to configure decision variables based on the regional situation set to construct a joint optimization model. It performs distributed sub-problem decomposition for the global joint optimization of the heterogeneous unmanned cluster. Through iterative solution of sub-problems, it dynamically adjusts the flight trajectory and task allocation plan of each unmanned device. Combined with conflict verification and control of edge nodes, it updates the resource allocation of each edge node to achieve collaborative control of each unmanned device.

[0011] The decision feedback module is used to continuously monitor the environmental obstacle data and status data of each unmanned device and perform local replanning.

[0012] Specifically, the steps for constructing a joint optimization model include:

[0013] Based on any edge node, identify and number the unmanned equipment within its jurisdiction;

[0014] Traverse the regional situation set corresponding to the edge node. For any unmanned device, call the corresponding path to be traveled. Combine the environmental obstacle data and take the discrete time series as the benchmark to obtain the three-dimensional coordinate position, velocity vector and obstacle adjustment parameters at each time point, and generate the trajectory variables configured for the unmanned device under the jurisdiction of the edge node.

[0015] Based on state data and load data, state variables and load variables are constructed using the timestamps of trajectory variables, and then the state variables and load variables are incorporated into the trajectory variables.

[0016] Based on the regional situation set, the task allocation ratio, allocated data volume, and transmission power of each unmanned device are obtained, and task variables are generated for the unmanned devices under the jurisdiction of the edge nodes.

[0017] Based on the trajectory variables and task variables of the same unmanned equipment, coupled variable pairs are constructed, and the coupled variable pairs of unmanned equipment under the jurisdiction of the edge node are integrated to form a set of decision variables for the area under the jurisdiction of the edge node;

[0018] Based on the regional situation set, a positive mapping and reverse traction are established to generate a set of coupling rules for path trajectory and task allocation.

[0019] Specifically, the steps for generating the coupling rule set between the path trajectory and task assignment include:

[0020] Based on the regional situation set and the three-dimensional coordinates of the unmanned equipment and the corresponding edge node, the straight-line distance between the unmanned equipment and the edge node is calculated, and the distance fit is calculated in combination with the effective service radius of the edge node.

[0021] Based on the regional situation set, acquire environmental adaptability, task priority and unmanned equipment status data, and calculate task gain;

[0022] Obtain a preset gain threshold and compare it with the task gain to perform a pre-allocation determination;

[0023] If the task gain is not less than the gain threshold and the remaining computing power of the edge node is greater than the amount of data to be allocated, then the edge node is determined to meet the task allocation conditions; otherwise, the task allocation ratio in the decision variable set is set to 0.

[0024] For edge nodes that meet the conditions for task allocation, obtain the task gain of unmanned equipment in the edge node and the remaining computing power of the edge node, and calculate the task ratio and computing power adaptation coefficient through normalization processing.

[0025] Based on task proportion and computing power adaptation coefficient, calculate the task allocation ratio of unmanned equipment at edge nodes;

[0026] Based on task requirements, an allocation threshold is set, and an edge node is determined to be a task allocation node for unmanned equipment only when the task allocation ratio is not less than the allocation threshold.

[0027] Set a trajectory approximation threshold for task allocation nodes and compare it with the straight-line distance to determine the adjustment rules for trajectory variables and construct a reverse traction between task allocation and the path trajectory;

[0028] The forward mapping and reverse traction are integrated into a set of coupling rules.

[0029] Specifically, the steps for constructing a joint optimization model also include:

[0030] Set motion constraints, communication constraints, resource constraints, energy constraints, and cutoff constraints to configure a set of coupled constraints for edge nodes;

[0031] Based on the decision variable set, the flight delay, transmission delay, computation delay, and result return delay of each unmanned device are obtained respectively;

[0032] The mission latency of the unmanned equipment is constructed by summing the latency, and the total energy consumption of the unmanned equipment is constructed based on the sum of flight energy consumption and communication energy consumption.

[0033] By weighted summing of task latency and total energy consumption, an objective function is constructed for the unmanned equipment;

[0034] Based on the objective function values ​​of all unmanned devices in the edge nodes, the edge objective function is constructed by summing the objective function values ​​of each unmanned device, and finally a joint optimization model is generated.

[0035] Specifically, the steps for distributing subproblems include:

[0036] Based on the unmanned equipment number managed by the edge node, extract the objective function corresponding to the unmanned equipment to set up independent sub-problems;

[0037] The solution boundaries of each subproblem are extracted from the set of coupling constraints, the associated parameters are labeled based on the set of coupling rules, and a subproblem mapping table is constructed.

[0038] Specifically, the iterative solution of the subproblems includes:

[0039] Based on the unmanned equipment under the jurisdiction of the edge node, initialize the corresponding coupling variable pairs and perform iterative initialization;

[0040] In each iteration, the task gain is calculated based on the forward mapping and determined in conjunction with the gain threshold;

[0041] When the edge node has the conditions for task allocation, update the task allocation ratio and reduce the transmission power;

[0042] When an edge node does not meet the task allocation conditions, the task allocation ratio of the current edge node is set to zero, and the task gain is switched to an edge node that meets the standard, while increasing the transmission power.

[0043] The straight-line distance between the unmanned device and the edge node is calculated based on the reverse traction calculation, and the determination is made in combination with the trajectory approximation threshold;

[0044] When the communication distance is insufficient, adjust the velocity vector and obstacle adjustment parameters;

[0045] When the communication distance meets the target, the original velocity vector is maintained, and the obstacle adjustment parameters are only updated when an obstacle is detected.

[0046] After the coupling rules are determined and corrected, the three-dimensional coordinates of the unmanned equipment are calculated, and the iteratively updated coupling variables are substituted into the constraint subset of the corresponding subproblem for constraint verification.

[0047] After the constraint verification is passed, the iteration difference is calculated based on the objective function values ​​in adjacent iterations;

[0048] The convergence threshold in the iterative initialization is used to determine the convergence of the iteration difference. The iteration stops only when the iteration difference is less than the convergence threshold or the number of iterations is greater than the number of iterations threshold, and the final pair of coupled variables is used as the initial optimal solution of the current subproblem.

[0049] Specifically, conflict verification and control at edge nodes include:

[0050] Integrate the initial optimal solutions of the unmanned devices under the jurisdiction of the edge nodes, and calculate the total task volume of each edge node;

[0051] For each edge node, based on resource constraints, the corresponding edge capacity limit is invoked;

[0052] Resource conflicts are determined to exist at edge nodes only when the total number of tasks exceeds the edge's capacity limit.

[0053] By utilizing communication constraints in the coupling constraint set, it is determined whether there is a conflict in the communication links of edge nodes; a link conflict is determined to exist only when the actual transmission rate is less than the minimum transmission rate.

[0054] Obtain the optimal trajectory position of all unmanned devices under the jurisdiction of the edge nodes. For any two unmanned devices A and B belonging to different edge nodes, calculate the cross-node distance at the same time using Euclidean distance.

[0055] The edge nodes to which unmanned devices A and B belong are determined to have trajectory conflicts only when the distance between nodes is less than the preset trajectory safety distance.

[0056] Specifically, conflict verification and control at edge nodes also include:

[0057] Integrate the conflict determination results of all edge nodes within the control area of ​​the heterogeneous unmanned cluster, and execute the edge computing tasks of the unmanned equipment under its jurisdiction for edge nodes that do not have any conflicts;

[0058] For edge nodes with any conflicts, they are marked as conflict nodes, and reconciliation instructions are sent to the corresponding conflict nodes to carry out collaborative reconciliation, including resource reconciliation, link reconciliation, and trajectory reconciliation.

[0059] For the optimal set of decision variables after coordination and harmonization, the marginal objective function values ​​of the corresponding marginal nodes are calculated using the joint optimization model;

[0060] Global constraint verification is performed based on the coupled constraint set, and the optimality of the marginal objective function is determined by comparing the marginal objective function values ​​before and after harmonicization.

[0061] If the global constraint verification is satisfied and the marginal objective function is optimal, then the reconciliation of the corresponding conflict node is valid; otherwise, conflict detection and reconciliation are repeated until the reconciliation of all conflict nodes is valid.

[0062] Specifically, the steps of coordination and reconciliation include:

[0063] For resource conflicts, the unmanned device with the lowest task priority in the edge node is extracted as the reconciliation device, and the task allocation ratio of the reconciliation device in the corresponding conflict node is reduced by a preset adjustment range.

[0064] Traverse the edge nodes that govern and harmonize the equipment and have no conflicts, and allocate the adjusted task allocation ratio to the edge nodes that meet the task gain target. The allocation is based on the principle of prioritizing high task allocation ratios.

[0065] Conflict checks are performed on the edge nodes that receive the task proportion. If there are no conflicts, the reconciliation is determined to be successful, and the corresponding decision variable set is updated. If the reconciliation is not successful, the adjusted task allocation proportion is allocated to the remaining edge nodes under its jurisdiction in turn until the reconciliation is successful.

[0066] If all edge nodes of the governing reconciliation equipment are in a state of reconciliation failure, then increase the adjustment range until reconciliation is achieved.

[0067] For link conflicts, traverse the unmanned devices under the jurisdiction of the conflicting node. If the task gain of the device and the corresponding edge node exceeds the limit, reduce the transmission power; otherwise, increase the transmission power.

[0068] After adjusting the transmission power, a conflict check is performed. If a conflict still exists after the adjustment, the adjustment ratio is corrected. If there is no conflict, the reconciliation is successful.

[0069] For trajectory conflicts, for unmanned devices A and B that are in conflict, based on their corresponding task priorities, the unmanned device with the lower task priority is used as the reconciliation device, and the velocity vector is adjusted by changing the direction of travel.

[0070] After adjustment, the trajectory conflict is re-verified. If the conflict still exists after adjustment, the velocity vector of another unmanned device is adjusted again. If there is no conflict, the reconciliation is successful.

[0071] Specifically, the steps of local replanning include:

[0072] The system acquires the real-time updated regional situation set as the real-time situation parameters, and simultaneously reads the baseline situation parameters and sets the detection period.

[0073] The change in situation is calculated based on the absolute difference between real-time situation parameters and baseline situation parameters.

[0074] Multidimensional real-time monitoring is performed by using preset detection thresholds and changes in the situation.

[0075] If the changes in the situation are all less than the corresponding detection threshold, it is determined that there is no significant change, and the decision variable set after conflict control is maintained.

[0076] If the change in situation is not less than the corresponding detection threshold, it is judged as a significant change, and the task allocation ratio of the unmanned equipment to the edge node is adjusted.

[0077] The beneficial effects of this invention are:

[0078] By constructing a regional situation set as a whole, integrating the status, tasks, environmental obstacle data of various unmanned devices, as well as the data carried by edge nodes and the list of managed devices, comprehensive data support is provided for the joint optimization of trajectory and task allocation, avoiding the separation optimization problem caused by data fragmentation in existing technologies. Based on this situation set, decision variables are configured and a joint optimization model is constructed. The flight trajectory, obstacle avoidance path, and task allocation plan of unmanned equipment are incorporated into the same optimization framework. Through distributed sub-problem decomposition and iterative solution, collaborative decision-making between the two is achieved, breaking the limitation of existing technologies that optimize them as independent problems. This allows changes in flight path to adapt to communication link quality and edge node access selection. At the same time, task load characteristics and priorities can also guide path planning in reverse, ensuring that path planning and task unloading are mutually adapted and optimally coordinated. Combined with edge node conflict verification and control, resource allocation is dynamically updated and unmanned equipment is collaboratively controlled, further strengthening the coupling adaptability between trajectory and task allocation. This avoids increased task processing latency and decreased loading efficiency due to path changes, or unreasonable path planning due to changes in task load. By continuously monitoring environmental obstacles and status data of unmanned equipment and performing local replanning, the system dynamically adapts to scene changes, maintains the optimal coupling state of trajectory and task allocation, effectively improves the overall system performance, solves the core pain point of non-optimal system performance under the existing separate optimization mode, and achieves efficient and stable operation of heterogeneous unmanned cluster full-domain collaborative control. Attached Figure Description

[0079] Figure 1 This is a structural diagram of a heterogeneous unmanned swarm collaborative control system based on edge computing.

[0080] Figure 2 This is a flowchart of the process for constructing the joint optimization model in this invention;

[0081] Figure 3 This is a flowchart illustrating the iterative solution of the subproblem in this invention;

[0082] Figure 4 This is a structural diagram of the jurisdiction area of ​​the edge node in this invention. Detailed Implementation

[0083] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0084] Example 1

[0085] refer to Figures 1 to 4 As shown in the figure, this embodiment introduces a heterogeneous unmanned swarm collaborative control system based on edge computing, including: a control module, a situational awareness module, a joint optimization module, and a decision feedback module;

[0086] The management and control module receives tasks issued by users to obtain corresponding task data, including task area, task type, task priority, deadline, data collection requirements, and computing power requirements. Through preliminary matching and planning of unmanned equipment, edge nodes, and travel paths, combined with the travel paths of the unmanned equipment, it matches the appropriate equipment type and associated nodes for the issued tasks, thereby constructing an initial planning table for the task area. The initial planning table contains the unmanned equipment to be used, the travel paths to be taken, and the edge nodes associated with each unmanned equipment during its travel. Based on the initial planning table, it generates initial scheduling instructions, including the travel paths of the unmanned equipment, the data to be collected, and the associated nodes of each unmanned equipment.

[0087] In this embodiment, in emergency rescue scenarios, due to the limited computing power of unmanned equipment, it cannot independently handle complex computing tasks. Cloud computing has long distances and high latency. In this case, edge nodes will support unmanned equipment to complete computing tasks that it cannot handle independently. The allocation of computing power, as the control bar of the formation, ensures the quality of the communication link. For example, deploying high-computing-power task equipment to the coverage area of ​​edge nodes ensures that the task gain meets the standard. Since the unmanned equipment needs to process computing tasks in real time during its journey from the current location to the task area, such as image pre-recognition and obstacle prediction, it needs to be supported by computing power provided by edge nodes along the path to avoid exceeding the task processing latency limit due to distance from edge nodes. Among them, edge nodes include ground base stations, ground edge gateways, and ground edge computing nodes.

[0088] The task types include search, accompanying support, intervention, relay, transportation, and monitoring, which correspond to the user's built-in operation templates. Based on the current position of the unmanned equipment and the center coordinates of the task area, the Dijkstra algorithm is used to plan the shortest path, thereby generating the path to be traveled from the unmanned equipment to the task area.

[0089] The initial matching and planning of unmanned equipment, edge nodes, and travel paths includes:

[0090] In response to receiving and issuing tasks, the system calls the user-pre-built task template library and matches the device type that is compatible with the issued task based on the matching logic of task type and device type. This device type is used as the adaptation type for this issued task. For example, search / monitoring / intervention tasks are adapted to aerial photography drones, environmental monitoring / search and rescue tasks are adapted to a combination of ground robots and aerial photography drones, and transportation / relay tasks are adapted to heavy-duty drones.

[0091] Traverse the idle unmanned devices in the heterogeneous unmanned cluster and filter the candidate device pool based on the adaptation type;

[0092] Based on constraints in the task data, such as task area, deadline, data collection density, and total computing power requirement, the device configuration scheme is generated through the recommendation terminal built into the control center.

[0093] Select the corresponding number of unmanned devices from the candidate device pool based on the device configuration scheme, and assign them sequential numbers in the format of unmanned device. , , The total number of collaborative devices. For example, if the matched collaborative devices include UAV A, UAV B, and ground search and rescue robot C, they will be numbered as Task Device 1, Task Device 2, and Task Device 3 in sequence.

[0094] For each collaborative device, based on the path to be traveled, the communication coverage distance of the edge nodes is used as the matching criterion, and all edge nodes within the task area and along the path to be traveled are used as the matching range. For each collaborative device, associated nodes are matched, that is, edge nodes within the task area are selected first, provided that the communication coverage distance within the path segment is greater than the straight-line distance between the device's current location and the node, the remaining computing power is greater than the computing power requirement of the device segment, and the task gain meets the standard.

[0095] Based on the collaborative device numbers and associated relationships, a preliminary planning table is constructed for issuing tasks, including collaborative device numbers, device types, paths to be taken, associated nodes, pre-allocated computing power ratios for nodes, and data collection requirements. Initial scheduling instructions are generated based on the preliminary planning table, specifying the path, data collection requirements, associated nodes, task priority execution rules, and cluster grouping management for each unmanned device.

[0096] The situational awareness module responds to the initial scheduling command and, throughout the entire process of the unmanned equipment (UAV) moving from its current location to the mission area, acquires status data, environmental obstacle data, and payload data based on the UAV's data acquisition terminal. It then combines this data with mission data to construct a raw dataset for each UAV. Based on the list of UAVs managed by the edge nodes, it integrates the raw datasets of all UAVs in the list, and simultaneously combines this with the edge nodes' carrying data to construct a regional situational awareness set for each edge node, serving as the data foundation for decomposing sub-problems. The status data includes the UAV's 3D coordinate position, flight speed, attitude angle, remaining battery power, and kinematic parameters; the mission data includes the amount of business data to be processed, mission type, mission priority, and mission latency requirements. The environmental obstacle data includes obstacle coordinates, dimensions, and safe obstacle avoidance distance. The list of unmanned devices under the jurisdiction of the edge nodes is generated based on the jurisdiction of the edge nodes and the real-time three-dimensional coordinates of all unmanned devices. It stores the unmanned device number, location, and affiliation of the corresponding edge node. The edge nodes are pre-planned and deployed by those skilled in the art based on the control area of ​​the heterogeneous unmanned cluster, and are fixedly deployed to key locations within the task coverage area. They have fixed communication coverage, computing resources, and access capabilities to provide edge computing services for unmanned devices. The data carried includes the communication coverage, total computing power, remaining computing power, task processing rate, maximum number of unmanned devices connected, communication channel capacity, and number of currently connected unmanned devices of the corresponding edge node.

[0097] The joint optimization module is used to configure decision variables based on the regional situation set to construct a joint optimization model. Using the joint optimization model as the solution object, the global joint optimization of the heterogeneous unmanned cluster is decomposed into sub-optimization problems of unmanned devices in each edge node. This is done through distributed sub-problem decomposition. By iteratively solving the sub-problems, the flight trajectory and task allocation plan of each unmanned device are dynamically adjusted. At the same time, combined with the conflict verification and control of the edge nodes, the resource allocation of each edge node is updated to enable collaborative control of each unmanned device, including the desired flight trajectory, task allocation strategy, and emergency obstacle avoidance rules. The collaborative control commands are then sent to the corresponding unmanned devices through the corresponding edge nodes.

[0098] The decision feedback module is used to continuously monitor the environmental obstacle data and status data of each unmanned device when the unmanned cluster executes collaborative control commands, and to perform local replanning to ensure that the system always maintains a collaborative optimization state in a dynamic environment.

[0099] Furthermore, the steps for constructing a joint optimization model include:

[0100] Based on any edge node, assign an identification number to the unmanned equipment within its jurisdiction. For example, the unmanned equipment number is... , , This represents the total number of unmanned devices under the jurisdiction of the edge nodes.

[0101] Traverse the regional situation set corresponding to the edge nodes. For any unmanned device, call the corresponding path to be traveled. Combined with environmental obstacle data, and based on discrete time series, represent the path to be traveled of the unmanned device as three-dimensional coordinate position, velocity vector, and obstacle adjustment parameters, and generate the edge node as the first node under its jurisdiction. Trajectory variables configured for unmanned equipment In the formula, , , The edge nodes are unmanned devices. Configuration, the first The three-dimensional coordinate position, velocity vector, and obstacle adjustment parameters at each moment. Plan the total duration for the task;

[0102] Based on state data and load data, state variables and load variables are constructed using the timestamps of trajectory variables, and then the state variables and load variables are incorporated into the trajectory variables.

[0103] Based on the task data of the regional situation set and the task computing data carried by the edge nodes, the task allocation of unmanned equipment is represented by the task allocation ratio, the amount of allocated data, and the transmission power. Edge nodes are then generated as the subordinate nodes. Task variables configured for unmanned equipment ;in, Assign a vector, and , For unmanned equipment In the Tasks are constantly being assigned to edge nodes. The task allocation ratio in this application is as follows: For computational tasks that are difficult for unmanned equipment to complete independently, they will be allocated to the corresponding edge nodes for collaborative processing according to the task allocation ratio. , This represents the total number of edge nodes. , The edge nodes (i.e., the scheduling entities) are unmanned devices. Configuration, the first The amount of data allocated and the transmission power at each moment;

[0104] In this embodiment, the edge node that governs the unmanned equipment is the scheduling entity; that is, the edge node among the unmanned equipment under the jurisdiction of the edge node is the scheduling entity, and the task is assigned to the edge node. The edge nodes in the process are unloading objects, including both the scheduling body and other adjacent cooperating edge nodes, and are optional receiving objects for task unloading;

[0105] By binding the trajectory variables and task variables configured for the same unmanned equipment, a new unmanned equipment can be formed. The coupled variable pairs, combined with the coupled variable pairs of the unmanned equipment under the jurisdiction of the edge node, form the decision variable set of the area under the jurisdiction of the edge node;

[0106] For example, a certain edge node manages a total of 3 unmanned devices, which are respectively unmanned device numbers. If the total task planning time is 2, then the corresponding discrete time is... ;

[0107] For number Taking drones as an example, The corresponding three-dimensional coordinate position, velocity vector, and obstacle adjustment parameters are as follows: , 0, The corresponding three-dimensional coordinate position, velocity vector, and obstacle adjustment parameters are as follows: , If the value is 0.8, then the trajectory variable is... ;

[0108] If the number of this edge node Adjacent collaborative edge node numbers , At this time, the task allocation ratio to this edge node is 1, and the task allocation ratio to adjacent cooperating edge nodes is 0. Simultaneously, the corresponding allocated data volume and transmission power are 10 and 2, respectively. When the task allocation ratio to this edge node is 0.2 and the task allocation ratio to adjacent collaborating edge nodes is 0.8, and the corresponding allocated data volume and transmission power are 15 and 3 respectively, then the task variable is... ;

[0109] Number at this time The coupling variable pairs of the drone are , combined , Coupled variables , Then the set of decision variables for this edge node is ;

[0110] Based on the regional situation set, a positive mapping and reverse traction are established to generate a set of coupling rules for path trajectory and task allocation;

[0111] Based on the type of unmanned equipment, upper limit thresholds are set for the acceleration, velocity, and turning radius of each unmanned equipment. At the same time, combined with the coordinates and dimensions of obstacles in the regional situational awareness, constraints are imposed on the three-dimensional coordinates in the trajectory variables. That is, the Euclidean distance between the coordinates of the unmanned equipment and the coordinates of the obstacle at any time is not less than the safe obstacle avoidance distance. In each round of trajectory variable update, the velocity, acceleration, and turning radius in the trajectory variables are restricted from not exceeding the corresponding upper limit thresholds, thereby constructing motion constraints. Among them, the upper limit thresholds are determined by the hardware parameters of the unmanned equipment and are directly read.

[0112] To ensure the reliability of data transmission during task allocation, the Shannon formula is used to constrain the minimum transmission rate between the unmanned device and its associated edge node. Specifically, the actual transmission rate between the unmanned device and its associated edge node is calculated using the Shannon formula, and the calculated actual transmission rate is not less than the preset minimum transmission rate, thus constructing a communication constraint. Among these constraints, the maximum data volume per frame in the task data and the maximum allowable latency of the task are obtained. The basic transmission rate is calculated using the ratio of the maximum data volume per frame to the maximum allowable latency. At the same time, the minimum stable rate supported by the communication standard is obtained, and the maximum value between the basic transmission rate and the minimum stable rate is used as the minimum transmission rate.

[0113] The total computing power of the edge nodes is obtained, and the discrete time step in the trajectory variables is combined with the product of the total computing power and the discrete time step to set the edge carrying capacity limit. For each unmanned device, the product of the task allocation ratio and the allocated data volume is calculated. Based on all unmanned devices in the edge nodes, the product is summed to calculate the total task. Resource constraints are set based on the total task and the edge carrying capacity limit, that is, the total task does not exceed the edge carrying capacity limit.

[0114] Based on the energy consumption terminal configured in the unmanned equipment, the flight energy consumption and communication energy consumption of the unmanned equipment are obtained. At the same time, the remaining energy of the unmanned equipment is obtained from the regional situational awareness, so as to constrain the sum of the flight energy consumption and communication energy consumption of the unmanned equipment to not exceed the remaining energy, thereby generating energy constraints.

[0115] Based on the task latency requirements of regional situational awareness, a cutoff constraint is set for the coupled variable pairs of unmanned equipment, that is, the end-to-end latency from data generation to distribution and processing completion is not greater than the cutoff time; wherein, the cutoff time is preset by those skilled in the art before collaborative control.

[0116] Integrate motion constraints, communication constraints, resource constraints, energy constraints, and cutoff constraints to configure a set of coupled constraints for edge nodes;

[0117] Based on the decision variable set, the flight delay, transmission delay, computation delay, and result feedback delay of each unmanned device are obtained. The task delay of the unmanned device is constructed by summing the delays, and the total energy consumption of the unmanned device is constructed based on the sum of flight energy consumption and communication energy consumption. Among them, the flight delay is obtained by the ratio of the position change of the unmanned device to the velocity vector, the transmission delay is obtained by multiplying the task allocation ratio and the allocated data volume and then dividing the transmission rate, and the computation delay is obtained by multiplying the task allocation ratio and the allocated data volume and then dividing the total computing power.

[0118] A target function is constructed for the unmanned equipment by weighted summation of task latency and total energy consumption. The weighting coefficients are set by those skilled in the art based on task requirements, and the sum of the weighting coefficients for task latency and total energy consumption is 1. For example, if the task scenario prioritizes low latency, such as emergency rescue or real-time intervention, the weighting coefficient for task latency ranges from 0.7 to 0.9, and the corresponding weighting coefficient for total energy consumption ranges from 0.1 to 0.3, prioritizing the timeliness of task completion. For scenarios prioritizing low energy consumption, such as long-term inspections and environmental monitoring, the weighting coefficient for task latency ranges from 0.1 to 0.3, while the corresponding weighting coefficient for total energy consumption ranges from 0.7 to 0.9, in order to extend the equipment's battery life. When the task scenario is in a balanced mode, such as material transportation and relay communication, the weighting coefficient for task latency ranges from 0.4 to 0.6, while the corresponding weighting coefficient for total energy consumption ranges from 0.4 to 0.6, in order to balance timeliness and battery life.

[0119] Based on the objective function values ​​of all unmanned devices in the edge node, an edge objective function is constructed for the edge node by summing the objective function values ​​of each unmanned device;

[0120] Using the set of decision variables as the solution carrier, combined with the set of coupling rules and the set of coupling constraints as solution restrictions, and the marginal objective function as the criterion for determining the optimal solution, a joint optimization model is generated.

[0121] Furthermore, the steps for generating the coupling rule set between path trajectories and task assignments include:

[0122] Based on the regional situation set, obtain the first Unmanned equipment at all times 3D coordinates and the coordinates of the edge nodes. Calculate the straight-line distance between the unmanned device and the edge node. Combining the effective service radius of the edge nodes, the ratio of the straight-line distance to the effective service radius is subtracted from 1 to obtain the adaptation ratio. The maximum value between 0 and the adaptation ratio is taken as the distance adaptation degree. If the distance fit is 0, it indicates that the device is outside the service range; if the distance fit is 1, it indicates that the unmanned device is directly above the edge node.

[0123] Based on the regional situation set, distance adaptability, task priority, and unmanned equipment status data are obtained, and the task gain is calculated as shown in the following expression:

[0124]

[0125] In the formula, For unmanned equipment With edge nodes Matching task gain, , , , These are weighting coefficients, set by those skilled in the art based on task requirements. For example, in low-latency scenarios, distance has the highest weight, with corresponding weighting coefficients of 0.4, 0.3, 0.2, and 0.1; in high-reliability scenarios, the environment has the highest weight, with corresponding weighting coefficients of 0.2, 0.2, 0.2, and 0.4; and in balanced scenarios, weight equilibrium has the corresponding weighting coefficients of 0.25, 0.25, 0.25, and 0.25. To determine the state fit, the remaining battery power of the unmanned device is obtained, and the ratio of the remaining battery power to the full battery power is used as the state fit. For normalized task priorities, For environmental adaptability;

[0126] Call the preset gain threshold Perform a pre-allocation judgment; if If the remaining computing power of the edge node is greater than the amount of data to be allocated by the unmanned equipment, then the edge node is deemed to meet the conditions for task allocation; if If the remaining computing power of the edge nodes is not greater than the amount of data to be allocated by the unmanned equipment, it indicates that task allocation is not satisfied at this time, and the task allocation ratio in the decision variable set will be adjusted. Setting it to 0 disables task assignment to this edge node; where the gain threshold... The value is preset by technicians according to the communication standard and mission reliability requirements, and the range is 1.5~2.5. For example, the gain threshold of 5G communication standard is 2.0 to meet the high reliability requirements, and the gain threshold of data transmission radio standard is 1.5 to meet the medium reliability requirements.

[0127] For edge nodes that meet the task allocation conditions, obtain the task gain of the unmanned equipment in all edge nodes. Through normalization processing, calculate the task ratio of the unmanned equipment to all edge nodes that meet the conditions. Simultaneously obtain the remaining computing power of each edge node. Through min-max normalization processing of the remaining computing power, obtain the computing power adaptation coefficient of each edge node.

[0128] The task allocation ratio of unmanned equipment in each edge node is calculated by multiplying the task ratio by the computing power adaptation coefficient.

[0129] Acquire real-world test data for heterogeneous unmanned cluster tasks, including task success rate, latency compliance rate, and computing power utilization under different allocation ratios. Using the ROC curve method, set an allocation threshold, such as 0.6. Only when the task allocation ratio is not less than the allocation threshold are edge nodes determined to be task allocation nodes for unmanned devices. At this time, the edge nodes It is an unmanned device Task allocation node;

[0130] Based on the task allocation results of edge nodes and unmanned equipment, a trajectory approximation threshold is set. For example, the trajectory approximation threshold ranges from 30 to 60m, and the effective service radius of the edge node is usually 50 to 100m. Among them, 30 to 60m is the core coverage area with stable links. The trajectory approximation threshold for Level 1 tasks is set to 30m, the trajectory approximation threshold for Level 2 tasks is set to 40m, and so on, in order to build a reverse traction of task allocation to the path trajectory.

[0131] when This indicates unmanned equipment With task allocation node If the distance is too far, the communication range will not meet the requirements, resulting in high communication and latency risks. Edge nodes need to adjust the velocity vector and obstacle adjustment parameters in the trajectory variables to guide the unmanned equipment to the task allocation node. Flight approaching; when This indicates that the unmanned equipment is within the stable communication range of the task allocation node. At this point, the communication distance meets the standard, and there is no need to perform trajectory approximation. It can simply fly normally according to the obstacle and motion constraints.

[0132] The operational logic and adjustment rules of forward mapping and reverse traction, which determine the rationality of task allocation ratio based on trajectory position and the adjustment direction of trajectory variables based on task allocation ratio, are integrated into a set of coupled rules.

[0133] Furthermore, the steps of distributed subproblem decomposition include:

[0134] Based on the unmanned equipment number managed by the edge node, the objective function corresponding to the unmanned equipment is extracted, and the optimization of the coupling variable pairs of a single unmanned equipment to minimize the corresponding objective function is taken as an independent sub-problem;

[0135] For each unmanned device's subproblem, a subset of constraints corresponding to the unmanned device is extracted from the coupling constraint set as the solution boundary of the subproblem. At the same time, the corresponding associated parameters are labeled in combination with the coupling rule set, including the positive mapping association between the task allocation ratio and the three-dimensional coordinates, and the negative traction association between the task allocation ratio and the velocity vector, to ensure that the coupling rules are always followed when solving the subproblem and that the trajectory and task allocation are not separated.

[0136] The sub-problems of all edge nodes are summarized, and a sub-problem mapping table is constructed in the format of edge node, unmanned equipment number, sub-problem objective, and sub-problem boundary. This clarifies the sub-problem affiliation of each device and the association between each sub-problem and the edge objective function.

[0137] Furthermore, the iterative solution of the subproblems includes:

[0138] Unmanned devices based on edge nodes Initialize the corresponding coupling variable pairs , where the initial trajectory variable In the middle, the initial three-dimensional coordinate position The initial position and initial velocity vector of the unmanned equipment. Set the initial obstacle-free adjustment and initial obstacle adjustment parameters for the rated vector of the unmanned equipment. A value of 0 indicates that no obstacles were detected in the initial state, and no obstacle avoidance offset is needed on the original path; this is the initial task variable. In the initial task allocation ratio Take a uniform distribution ratio, such as Initial data allocation Determine the initial task data volume and initial transmission power. Taking the rated transmission power, for example, if the unmanned equipment is initially deployed in the area coordinates... ,but The corresponding rated flight speed is 10 m / s, and the trajectory is along the positive x-axis. If there are a total of 5 edge nodes in the region, that is ,but Each edge node is allocated 20% of the tasks, with an initial task data volume of 1000MB. If the rated transmission power is 5W, then ;

[0139] Those skilled in the art, based on experimental fitting, set the current iteration number. The system initializes the iterations by reading a convergence flag and reading a preset iteration count threshold and convergence threshold. For example, in the first iteration, the system sets... The threshold for the number of iterations is 50, and the threshold for convergence is 10. -3 ;

[0140] In each iteration, the task gain in the current iteration is calculated based on the forward mapping and judged in combination with the gain threshold. When the edge node has the conditions for task allocation, the task allocation ratio is updated with a preset adjustment step size. At the same time, the transmission power is reduced based on the preset adjustment ratio to save energy.

[0141] When an edge node does not meet the task allocation conditions, the task allocation ratio of the current edge node is set to zero, the task gain is switched to the edge node that meets the standard, the task allocation ratio is updated, and the transmission power is increased based on the adjusted ratio to ensure communication quality.

[0142] For example, in the first iteration, the coordinates of the edge node are... The corresponding straight-line distance is When the channel gain is 10 5 If the ambient occlusion factor is 1.2, then the task gain at this time... If the set gain threshold is 2, and the task allocation condition is determined based on the task gain, then... At this point, the edge nodes meet the task allocation conditions. When the step size is less than 0.02, the convergence speed is too slow and the number of iterations is too large. When the step size is greater than 0.1, the accuracy of the solution is insufficient and oscillations are prone to occur. In order to balance the convergence speed and the accuracy of the solution, the step size is adjusted. Update the task allocation ratio, i.e. Meanwhile, in order to balance link compensation effectiveness and energy consumption control, the transmission power is reduced by a 5% adjustment ratio, i.e. Specifically, to increase transmission power, the transmission power will be increased by a 5% adjustment ratio, i.e. Among them, the environmental occlusion coefficient is obtained from the environmental parameters of the regional situation concentration, which is the degree of environmental occlusion measured on site, and the value range is 1-2, where 1 is no occlusion and 2 is complete occlusion.

[0143] The straight-line distance in the current iteration is calculated based on the reverse traction calculation, and the judgment is made in combination with the trajectory approximation threshold. When the communication distance is not up to standard, the velocity vector is adjusted by a preset speed adjustment ratio. If the ratio is less than 5%, the approximation speed is too slow and cannot quickly reduce the risk of delay. If the ratio is greater than 15%, it exceeds the upper limit of the unmanned equipment's maneuverability and is prone to flight instability. To balance the approximation speed and flight safety, the speed adjustment ratio is set to 10%. The adjustment range is 10% of the rated speed. The updated velocity vector is obtained by adding the velocity vector output from the previous iteration. At the same time, the obstacle adjustment parameters are updated with the obstacle avoidance adjustment step size to optimize the obstacle avoidance path and approach the node.

[0144] For example, in one iteration, At this point, the edge node is the task allocation node, and the straight-line distance is 60m, while the trajectory approximation threshold is 50m. The communication distance does not meet the standard. Simultaneously, when it is necessary to adjust the obstacle avoidance path to approach the node, the step size is adjusted based on obstacle avoidance. Update obstacle adjustment parameters, i.e. If the value is updated from 0 to 0.1, it means that the path is offset by 0.1m towards the node, the obstacle avoidance path is optimized, and the unmanned equipment is pulled closer to the node. When there are no obstacles and no need to approach, the parameter value remains unchanged.

[0145] When the communication distance meets the target, the original velocity vector is maintained, and the obstacle adjustment parameters are updated only when an obstacle is detected, using the obstacle avoidance adjustment step size; otherwise, the original obstacle adjustment parameters are maintained.

[0146] After the coupling rules are determined and corrected, the three-dimensional coordinates of the unmanned equipment in the current iteration are calculated by combining the velocity vector and discrete time step of the current iteration with the position of the unmanned equipment in the previous iteration. The updated coupling variables are then substituted into the constraint subset of the corresponding subproblem for constraint verification. The next objective function is calculated only if all constraint verifications pass. Otherwise, the parameters that do not meet the constraints are corrected and re-verified until all constraints are met.

[0147] After the constraint verification is passed, the objective function value of the current iteration subproblem is calculated. Combined with the objective function value of the previous iteration, the iteration difference is calculated by the absolute difference. Convergence is determined by combining the convergence threshold. The iteration stops only when the iteration difference is less than the convergence threshold or the number of iterations is greater than the number of iterations threshold. The final pair of coupled variables is used as the initial optimal solution of the current subproblem. Otherwise, the iteration continues.

[0148] Furthermore, conflict verification and control at edge nodes include:

[0149] By integrating the initial optimal solutions of the unmanned devices under the jurisdiction of the edge nodes, and utilizing the resource constraints in the coupling constraint set, the total task volume of each edge node is calculated. ;

[0150] For each edge node, call the corresponding edge capacity limit. such as edge nodes The total computing power is 1000MB / s, and based on the unmanned equipment's task sampling period of 1.25Hz (collecting state data once every 0.8s), the discrete time step is set to 0.8, which corresponds to the upper limit of edge throughput. ;

[0151] when At this time, the edge nodes The system is capable of handling the computational workload of the unmanned equipment under its jurisdiction, and resources are determined to be conflict-free; if At this time, the edge nodes If the task cannot be carried out, it is judged as a resource conflict, and the corresponding edge node and all unmanned devices that have assigned tasks to this conflicting node (i.e., all unmanned devices under the jurisdiction of this edge node) are recorded to form a computing power conflict package.

[0152] Using communication constraints in the coupling constraint set, it is determined whether there is a conflict in the communication links of the edge nodes. When the actual transmission rate of an edge node is less than the minimum transmission rate, it is determined that there is a link conflict. The corresponding conflicting link, the unmanned equipment using the link, and the edge node to which it belongs are recorded, and the link conflict packet is visualized. If the actual transmission rate of an edge node is not less than the minimum transmission rate, it is determined that there is no link conflict.

[0153] Obtain the optimal trajectory position of all unmanned devices under the jurisdiction of edge nodes. For any two unmanned devices A and B belonging to different edge nodes, calculate the cross-node distance at the same time using Euclidean distance, and determine the cross-node collision by combining the preset trajectory safety distance. The optimal trajectory position is obtained from the initial optimal solution, and the trajectory safety distance is set by those skilled in the art, such as 10m.

[0154] If the distance across nodes is less than the trajectory safety distance, it indicates that there is a collision between unmanned devices A and B, which is determined to be a trajectory conflict. The corresponding conflicting device, its associated edge node, and the corresponding time are recorded to form a trajectory conflict packet. If the distance across nodes is not less than the trajectory safety distance, it is determined that there is no trajectory conflict.

[0155] The conflict determination results of all edge nodes within the control area of ​​the heterogeneous unmanned cluster are integrated. For edge nodes that do not have any conflicts, the task that the unmanned equipment under its jurisdiction needs to calculate at the edge is executed, and the calculation result is sent to the corresponding unmanned equipment after the calculation is completed.

[0156] For edge nodes with any conflicts, they are marked as conflict nodes, and reconciliation instructions are sent to the corresponding conflict nodes to carry out collaborative reconciliation, including resource reconciliation, link reconciliation, and trajectory reconciliation.

[0157] For the optimal solution after co-harmonization, i.e. the corresponding optimal decision variable set, the marginal objective function value of the corresponding marginal node is calculated using the joint optimization model. Global constraint verification is performed based on the coupling constraint set. At the same time, the marginal objective function values ​​before and after harmonization are compared to ensure that the marginal objective function value after harmonization is less than the marginal objective function value before harmonization, so as to make the marginal objective function optimal.

[0158] If the global constraint verification is satisfied and the marginal objective function is optimal, the reconciliation of the corresponding conflict node is valid, and a confirmation command is issued to proceed with the reconciliation. If the global constraint verification is not satisfied or the marginal objective function is not optimal, conflict detection and reconciliation are repeated until the reconciliation of all conflict nodes is valid.

[0159] Furthermore, the steps for coordinated reconciliation include:

[0160] For resource conflicts, the unmanned device with the lowest task priority among the edge nodes is extracted as a mediator. The task allocation ratio of the mediator to the corresponding conflicting node is reduced by a preset adjustment range, such as 0.1, for example, from 0.6 to 0.5. The task priority is presented in numerical form and set by technical personnel, such as priority values ​​1-5, where 1 is the lowest and 5 is the highest. The adjustment range is the product of the current state deviation and the adjustment upper limit, which is obtained through iterative convergence tests under different time lengths.

[0161] The process iterates through the edge nodes that govern the reconciliation device and have no conflicts, allocating the adjusted task allocation ratios to edge nodes with sufficient task gains. Allocation prioritizes nodes with higher task allocation ratios. For example, the edge nodes governing the reconciliation device include... , , The corresponding task allocation ratios are 0.2, 0.1, and 0.1, respectively. The adjusted task allocation ratio of 0.1 will then be allocated to the edge nodes. ;

[0162] Conflict checks are performed on the edge nodes that receive the task proportion. If there are no conflicts, the reconciliation is determined to be successful, and the corresponding decision variable set is updated. If the reconciliation is not successful, the adjusted task allocation proportion is allocated to the remaining edge nodes under its jurisdiction in turn. Conflict checks are performed once for each allocation until the reconciliation is successful.

[0163] If all edge nodes of the reconciliation equipment fail to reconcile, the adjustment range is increased, such as from 0.1 to 0.2, which means the task allocation ratio is reduced from 0.6 to 0.4 until reconciliation is successful.

[0164] For link conflicts, traverse the unmanned devices under the jurisdiction of the conflicting node. If the task gain of the device and the corresponding edge node is high, then... If the transmission power is reduced by 10%, for example, from 5W to 4.5W; if the task gain is low, i.e. This would increase the transmission power by 10%, such as from 5W to 5.5W, to balance the link load.

[0165] After adjusting the transmission power, a conflict check is performed. If a conflict still exists after the adjustment, the adjustment ratio is corrected. For example, if the transmission power is reduced by 10%, it is changed to a reduction of 15%. If there is no conflict, the reconciliation is successful.

[0166] For trajectory conflicts involving conflicting unmanned devices A and B, based on their respective task priorities, the unmanned device with the lower task priority is chosen as the reconciling device. For example, if device A has a task priority of 3 and device B has a task priority of 2, device B is chosen as the reconciling device. The velocity vector is adjusted by changing the direction of travel, for example, by an adjustment of 30°. At this time, the velocity vector changes from the positive x-axis direction. Adjusted to be along the positive x-axis and positive y-axis Meanwhile, the obstacle avoidance parameters are adjusted, such as by increasing them by 0.1, to optimize the obstacle avoidance path.

[0167] After adjustment, the trajectory conflict is re-verified. If the conflict still exists after adjustment, the velocity vector of another unmanned device is adjusted again. If there is no conflict, the reconciliation is successful.

[0168] Furthermore, the steps of local replanning include:

[0169] After completing the conflict verification and control, the real-time updated regional situation set is obtained as the real-time situation parameters. At the same time, the baseline situation parameters, including the baseline temperature, baseline task data volume, baseline load, and baseline task gain, are read out, and the detection cycle is set. The baseline situation parameters and the detection cycle are set by those skilled in the art.

[0170] The change in situation is calculated based on the absolute difference between real-time situation parameters and baseline situation parameters. For example, for ambient temperature, the temperature change is calculated based on the absolute difference between the real-time temperature of the regional situation and the baseline temperature. The change in situation also includes the change in task, load, and gain.

[0171] Multi-dimensional real-time monitoring is performed using preset detection thresholds and situational change parameters. If all situational change parameters are less than the corresponding detection thresholds, it is determined that there is no significant change, and the globally optimal decision variable set after conflict control is maintained. The original plan continues to be executed, and the next detection cycle begins. If there are situational change parameters that are not less than the corresponding detection thresholds, it is determined that there is a significant change, including significant changes in temperature, data volume, computing power load, and link quality. The task allocation ratio of the unmanned equipment to edge nodes is adjusted accordingly. For example, when the temperature changes significantly, the task allocation ratio increases by 0.1. When the data volume changes significantly, the task data volume increases, and the task allocation ratio to edge nodes is increased by 0.1, prioritizing allocation to computing power. Sufficient edge nodes; if the amount of task data decreases, reduce the proportion of tasks allocated to edge nodes by 0.1 to alleviate the load on edge nodes. When the computing power load changes significantly, the task allocation ratio decreases by 0.1. When the link quality changes significantly, if the task gain decreases, increase the power by 10%; if the task gain increases, decrease the power by 10%. The detection threshold is obtained through experimental fitting. Based on the task execution results under different temperatures, different amounts of task data, different computing power loads, and different task gains, a historical dataset is constructed. For each situational parameter, all parameter changes that cause the task latency to exceed the cutoff constraint by 10% are statistically analyzed, and the 95th percentile of the change is taken as the detection threshold of the corresponding parameter.

[0172] Working principle and effects:

[0173] By acquiring status, task, and environmental obstacle data for each unmanned device, and combining this with edge node data and a list of managed devices, a regional situation set is constructed. This provides comprehensive data support for subsequent joint optimization, resolving the fragmented optimization problem caused by data isolation in existing technologies. Based on this situation set, decision variables are configured, and a joint optimization model is built. The global joint optimization of the heterogeneous unmanned cluster is decomposed into distributed sub-problems. Through iterative solutions, the flight trajectories and task allocation plans of each unmanned device are dynamically adjusted, breaking the limitations of independent optimization and achieving strongly coupled collaborative decision-making, thus improving the overall system efficiency. Combined with edge node conflict verification and resource allocation updates, collaborative control of unmanned devices is achieved, preventing path changes from affecting task processing efficiency or causing unreasonable paths due to task load. Continuous monitoring of environmental obstacles and status data of unmanned devices and local replanning are performed to dynamically adapt to scene changes, maintain optimal coupling between trajectory and task allocation, and ensure stable and efficient system operation.

[0174] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A heterogeneous unmanned swarm collaborative control system based on edge computing, characterized in that, include: Control module, situation awareness module, joint optimization module, decision feedback module; The control module is used to acquire the corresponding task data, and through the initial matching and planning of unmanned equipment, edge nodes and travel paths, combined with the travel path of the unmanned equipment, match the appropriate equipment type and associated nodes to generate the initial scheduling instruction. The situational awareness module responds to the initial scheduling command, acquires status data and environmental obstacle data, and combines the carrying data of the edge nodes and the list of unmanned devices under its jurisdiction to construct a regional situational awareness set for the edge nodes. The joint optimization module is used to configure decision variables based on the regional situation set to construct a joint optimization model. It performs distributed sub-problem decomposition for the global joint optimization of the heterogeneous unmanned cluster. Through iterative solution of sub-problems, it dynamically adjusts the flight trajectory and task allocation plan of each unmanned device. Combined with conflict verification and control of edge nodes, it updates the resource allocation of each edge node to achieve collaborative control of each unmanned device. The decision feedback module is used to continuously monitor the environmental obstacle data and status data of each unmanned device, and to perform local replanning. The steps for constructing the joint optimization model include: Identify and number the unmanned equipment within the jurisdiction of the edge nodes; Traverse the regional situation set corresponding to the edge node, call the path to be traveled by the unmanned equipment, and obtain the three-dimensional coordinate position, velocity vector, and obstacle adjustment parameters at each time point to generate the trajectory variables configured for the unmanned equipment under the jurisdiction of the edge node; Based on state data and load data, state variables and load variables are constructed using the timestamps of trajectory variables and then incorporated into the trajectory variables. Obtain the task allocation ratio, allocated data volume, and transmission power of the regional situation-centralized unmanned equipment, and generate task variables for the unmanned equipment under the jurisdiction of the edge nodes; Based on the trajectory variables and task variables of the same unmanned equipment, coupled variable pairs are constructed, and the coupled variable pairs of unmanned equipment under the jurisdiction of the edge node are integrated to form the decision variable set of the edge node; Based on the regional situation set, the rationality of the task allocation ratio is determined by the trajectory position, and the adjustment direction of trajectory variables is determined by the task allocation ratio, thereby generating a set of coupling rules between the movement trajectory and the task allocation.

2. The heterogeneous unmanned swarm collaborative control system based on edge computing according to claim 1, characterized in that, The steps for generating the coupling rule set between path trajectory and task assignment include: Calculate the straight-line distance between the unmanned equipment and the edge node, and combine it with the effective service radius of the edge node to calculate the distance fit. Acquire environmental adaptability, task priority, and unmanned equipment status data, and calculate task gain; By comparing the gain threshold with the task gain, a pre-allocation judgment is made to filter out edge nodes that meet the task allocation conditions, and the task ratio and computing power adaptation coefficient are calculated. Based on task proportion and computing power adaptation coefficient, calculate the task allocation ratio of unmanned equipment at edge nodes; Based on task requirements, an allocation threshold is set, and an edge node is determined to be a task allocation node for unmanned equipment only when the task allocation ratio is not less than the allocation threshold. A trajectory approximation threshold is set for the task allocation node and compared with the straight-line distance to determine the adjustment rules for the trajectory variables. When the straight-line distance > the trajectory approximation threshold, the velocity vector and obstacle adjustment parameters in the trajectory variables are adjusted to guide the unmanned equipment toward the task allocation node. During flight approach, if the straight-line distance is less than or equal to the trajectory approach threshold, normal flight is carried out according to obstacle and motion constraints. The rationality of task allocation ratio determined by trajectory position and the direction of trajectory variable adjustment determined by task allocation ratio are integrated into a coupled rule set.

3. A heterogeneous unmanned swarm collaborative control system based on edge computing according to claim 2, characterized in that, The steps for constructing a joint optimization model also include: Set motion constraints, communication constraints, resource constraints, energy constraints, and cutoff constraints to configure a set of coupled constraints for edge nodes; Based on the decision variable set, the flight delay, transmission delay, computation delay, and result return delay of each unmanned device are obtained respectively; The mission latency of the unmanned equipment is constructed by summing the latency, and the total energy consumption of the unmanned equipment is constructed based on the sum of flight energy consumption and communication energy consumption. By weighted summing of task latency and total energy consumption, an objective function is constructed for the unmanned equipment; Based on the objective function values ​​of all unmanned devices in the edge nodes, the edge objective function is constructed by summing the objective function values ​​of each unmanned device, and finally a joint optimization model is generated.

4. A heterogeneous unmanned swarm collaborative control system based on edge computing according to claim 3, characterized in that, The steps of distributed subproblem decomposition include: Based on the unmanned equipment number managed by the edge node, extract the objective function corresponding to the unmanned equipment to set up independent sub-problems; The solution boundaries of each subproblem are extracted from the set of coupling constraints, the associated parameters are labeled based on the set of coupling rules, and a subproblem mapping table is constructed.

5. A heterogeneous unmanned swarm collaborative control system based on edge computing according to claim 4, characterized in that, Iterative solutions to subproblems include: Initialize the coupling variable pairs and calculate the task gain in each iteration; When the edge node has the conditions for task allocation, update the task allocation ratio and reduce the transmission power; When an edge node does not meet the task allocation conditions, the task allocation ratio of the current edge node is set to zero, and the task gain is switched to an edge node that meets the standard, while increasing the transmission power. Calculate the straight-line distance between the unmanned device and the edge node; When the communication distance is insufficient, adjust the velocity vector and obstacle adjustment parameters; When the communication distance meets the target, the original velocity vector is maintained, and the obstacle adjustment parameters are only updated when an obstacle is detected. Calculate the three-dimensional coordinates of the unmanned equipment, and substitute the iteratively updated coupling variables into the constraint subset of the corresponding subproblem for constraint verification. After the constraint verification is passed, calculate the iterative difference. After stopping the iteration, the final pair of coupled variables is used as the initial optimal solution for the current subproblem.

6. A heterogeneous unmanned swarm collaborative control system based on edge computing according to claim 5, characterized in that, Conflict verification and control at edge nodes include: Integrate the initial optimal solutions of the unmanned devices under the jurisdiction of the edge nodes, and calculate the total task volume of each edge node; For each edge node, based on resource constraints, the corresponding edge capacity limit is invoked; Resource conflicts are determined to exist at edge nodes only when the total number of tasks exceeds the edge's capacity limit. By utilizing communication constraints in the coupling constraint set, it is determined whether there is a conflict in the communication links of edge nodes; a link conflict is determined to exist only when the actual transmission rate is less than the minimum transmission rate. Obtain the optimal trajectory position of all unmanned devices under the jurisdiction of the edge nodes. For any two unmanned devices A and B belonging to different edge nodes, calculate the cross-node distance at the same time using Euclidean distance. The edge nodes to which unmanned devices A and B belong are determined to have trajectory conflicts only when the distance between nodes is less than the preset trajectory safety distance.

7. A heterogeneous unmanned swarm collaborative control system based on edge computing according to claim 6, characterized in that, Conflict verification and control at edge nodes also include: Integrate the conflict determination results of all edge nodes within the control area of ​​the heterogeneous unmanned cluster, and execute the edge computing tasks of the unmanned equipment under its jurisdiction for edge nodes that do not have any conflicts; For edge nodes with any conflicts, they are marked as conflict nodes, and reconciliation instructions are sent to the corresponding conflict nodes to carry out collaborative reconciliation, including resource reconciliation, link reconciliation, and trajectory reconciliation. For the optimal set of decision variables after coordination and harmonization, the marginal objective function values ​​of the corresponding marginal nodes are calculated using the joint optimization model; Global constraint verification is performed based on the coupled constraint set, and the optimality of the marginal objective function is determined by comparing the marginal objective function values ​​before and after harmonicization. If the global constraint verification is satisfied and the marginal objective function is optimal, then the reconciliation of the corresponding conflict node is valid; otherwise, conflict detection and reconciliation are repeated until the reconciliation of all conflict nodes is valid.

8. A heterogeneous unmanned swarm collaborative control system based on edge computing according to claim 7, characterized in that, The steps of coordinated reconciliation include: Set up a reconciliation device for resource conflicts and reduce the task allocation ratio of the reconciliation device; Traverse the edge nodes that govern and coordinate equipment and have no conflicts, and allocate and verify the latest task allocation ratio based on the principle of prioritizing high task allocation ratios. If there are no conflicts, the reconciliation is determined and the corresponding decision variable set is updated; otherwise, the task allocation ratio is reallocated or the adjustment range is increased until the reconciliation is achieved. For link conflicts, traverse the unmanned equipment under the jurisdiction of the conflicting node, adjust the transmission power according to the task gain, and perform conflict verification. If the conflict still exists after adjustment, adjust the adjustment ratio; otherwise, the reconciliation is successful. For trajectory conflicts, adjust the velocity vector of the unmanned equipment with lower task priority and perform conflict verification. If the conflict still exists after adjustment, adjust the velocity vector of the other unmanned equipment; otherwise, the reconciliation is successful.

9. A heterogeneous unmanned swarm collaborative control system based on edge computing according to claim 8, characterized in that, The steps of local replanning include: The system acquires the real-time updated regional situation set as the real-time situation parameters, and simultaneously reads the baseline situation parameters and sets the detection period. The change in situation is calculated based on the absolute difference between real-time situation parameters and baseline situation parameters. Multidimensional real-time monitoring is performed by using preset detection thresholds and changes in the situation. If the changes in the situation are all less than the corresponding detection threshold, it is determined that there is no significant change, and the decision variable set after conflict control is maintained. If the change in situation is not less than the corresponding detection threshold, it is judged as a significant change, and the task allocation ratio of the unmanned equipment to the edge node is adjusted.

Images