Multi-agent collaborative operation control method and system based on wireless deterministic network

Abstract

The application relates to the technical field of multi-agent cooperative control, and discloses a multi-agent cooperative operation control method and system based on a wireless deterministic network. The system comprises a time synchronization calibration module, which collects initial time stamps and network transmission delay data of each agent node, sets synchronization protocol attributes for calibration, and establishes a cooperative operation network model; a task instruction distribution module, which associates operation task requirements with agent capability data, generates a cooperative operation instruction set through task decomposition; a path conflict detection module, which optimizes a cooperative optimization path based on network model initialization parameters; a cooperative state sensing module, which acquires real-time state data, analyzes and determines control strategy parameters; and a control parameter optimization module, which simulates verification and feeds back optimized parameters. The system improves the precision, efficiency and reliability of multi-agent cooperative operation and is suitable for various complex operation scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-agent collaborative control, and specifically to a multi-agent collaborative operation control method and system based on a wireless deterministic network. Background Art

[0002] With the rapid development of industrial automation, intelligent logistics, smart cities, and other fields, multi-agent collaborative operations have become an important means of handling complex tasks due to their efficiency, flexibility, and scalability. Multi-agent systems, through the coordinated cooperation of multiple agent nodes, can accomplish large-scale tasks that are difficult for a single agent to handle, such as multi-robot collaborative handling in smart warehousing, collaborative assembly of multiple devices on industrial production lines, and collaborative inspection by drone swarms. However, in practical applications, multi-agent collaborative operations face numerous challenges due to the uncertainties of wireless communications.

[0003] Traditional wireless communication networks (such as Wi-Fi and Bluetooth) suffer from large fluctuations in transmission delays, low time synchronization accuracy, and insufficient data transmission reliability. This makes it difficult to achieve high-precision synchronization in collaborative control among multiple agents. For example, in a multi-robot collaborative operation scenario, if the timestamps of each robot are not synchronized, it may cause deviations in action execution, leading to path conflicts or reduced work efficiency. The uncertainty of network transmission delays will cause lags in the distribution of task instructions, affecting the speed at which the agents respond to dynamic working environments. In addition, as the number of agents increases, the network topology becomes increasingly complex, and problems such as path planning conflicts, uneven task distribution, and poor adaptability of control parameters become more prominent, seriously restricting the stability and efficiency of multi-agent collaborative operations.

[0004] Existing research on multi-agent collaborative control has largely focused on wired network environments or idealized wireless communication models, making it difficult to adapt to the dynamic changes in actual wireless environments. While some wireless-based multi-agent control systems have introduced time synchronization mechanisms, synchronization accuracy is significantly affected by network latency fluctuations, and they lack the ability to dynamically adapt to node topology relationships. Traditional approaches to task instruction distribution often employ fixed task allocation strategies, failing to dynamically adjust to the real-time capabilities of the agents and operational requirements, resulting in low resource utilization. Path conflict detection and optimization algorithms suffer from slow convergence and a dependence of optimization results on initial parameters, making them difficult to address the multi-path collaboration requirements of complex operational scenarios.

[0005] Multi-agent collaborative operations lack a closed-loop feedback mechanism for state perception and control parameter optimization. Once set, control strategy parameters remain fixed and cannot be dynamically adjusted based on operational performance. This results in insufficient robustness in the face of emergencies (such as agent failures and network transmission interruptions). Therefore, building a multi-agent collaborative operation control system based on a wireless deterministic network that achieves high-precision time synchronization, dynamic adaptation of task instructions, intelligent optimization of path conflicts, real-time perception of collaborative states, and closed-loop optimization of control parameters has become a pressing technical challenge. Summary of the Invention

[0006] The purpose of the present invention is to provide a multi-agent collaborative task control method and system based on a wireless deterministic network to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides a multi-agent collaborative operation control system based on a wireless deterministic network, the system comprising:

[0008] A time synchronization calibration module, which is used to collect initial timestamp data and network transmission delay data of each intelligent node in the wireless deterministic network, set synchronization protocol attributes, perform synchronization calibration processing on the initial timestamp data and network transmission delay data based on the synchronization protocol attributes, and establish a multi-agent collaborative operation network model;

[0009] A task instruction distribution module is used to obtain task requirement data and agent capability data through information analysis technology, perform instruction conversion processing on the task requirement data and agent capability data using a task decomposition method, and generate a multi-agent collaborative task instruction set;

[0010] A path conflict detection module, the path conflict detection module is used to initialize conflict detection parameters based on the multi-agent collaborative operation network model, perform agent path optimization on the conflict detection parameters according to the collaborative operation goal, and iteratively obtain the multi-agent collaborative optimization path;

[0011] A collaborative state perception module, which is used to obtain real-time state data of each agent according to the multi-agent collaborative operation instruction set, perform state correlation analysis based on the real-time state data and the multi-agent collaborative optimization path, and determine the multi-agent collaborative operation control strategy parameters;

[0012] A control parameter optimization module is used to simulate and verify the multi-agent collaborative operation control strategy parameters, obtain collaborative operation control effect data, and perform feedback optimization on the multi-agent collaborative operation control strategy parameters based on the collaborative operation control effect data.

[0013] Preferably, the establishment of a multi-agent collaborative operation network model includes:

[0014] Determining network node identification attributes and node communication link attributes based on the synchronization protocol attributes;

[0015] Perform node identification and attribute labeling on the initial timestamp data and the network transmission delay data based on the network node identification attribute to obtain an agent node set and a node attribute set;

[0016] Performing connection relationship analysis on the set of intelligent agent nodes using the initial timestamp data and the network transmission delay data to construct a node adjacency connection relationship set;

[0017] Performing network abstraction on the node adjacency connection relationship set based on the node communication link attributes to obtain a node topology adjacency matrix;

[0018] The agent node set is topologically identified and connected based on the node topology adjacency matrix and the node attribute set to establish the multi-agent collaborative operation network model.

[0019] Preferably, the generating of a multi-agent collaborative operation instruction set includes:

[0020] Use the task decomposition method to perform task splitting and capability matching on the job task requirement data and the agent capability data to obtain a job subtask dataset and an agent adaptation dataset;

[0021] Performing instruction conversion processing on the job subtask data set and the agent adaptation data set respectively through a rule engine to obtain a subtask instruction set and an agent adaptation instruction set;

[0022] Merging and integrating the subtask instruction set and the agent adaptation instruction set to generate an initial collaborative operation instruction set;

[0023] An instruction verifier is used to verify the compliance of the initial collaborative operation instruction set to obtain the multi-agent collaborative operation instruction set.

[0024] Preferably, the iterative acquisition of the multi-agent collaborative optimization path includes:

[0025] Extracting conflict assessment indicators for the collaborative operation target to obtain a conflict assessment indicator set, and constructing a path optimization effect evaluation function based on the conflict assessment indicator set;

[0026] Determine the detection particle position and the detection particle speed according to the conflict detection parameters, and use the path optimization effect evaluation function to evaluate the effect of the conflict detection parameters to obtain a detection particle effect set;

[0027] Iteratively updating the detection particle position and the detection particle velocity based on the detection particle effect set until a preset stop condition is met, and screening and determining the detection particle with the best effect, wherein the best detection particle includes the intelligent agent path optimization position;

[0028] Based on the agent path optimization position, the path of the multi-agent collaborative operation network model is updated and adjusted to obtain the multi-agent collaborative optimization path.

[0029] Preferably, the determining of the multi-agent collaborative operation control strategy parameters includes:

[0030] Performing a network traffic distribution impact analysis based on the multi-agent collaborative optimization path to obtain network traffic distribution impact parameters, wherein the network traffic distribution impact parameters include path load impact parameters and collaborative operation impact parameters;

[0031] Constructing a collaborative control strategy library, and performing matching screening with the collaborative control strategy library based on the real-time status data and the network traffic distribution influencing parameters to obtain a multi-agent collaborative matching control strategy;

[0032] Based on the multi-agent collaborative matching control strategy, the real-time status data and the network traffic distribution influencing parameters are divided into threshold ranges to obtain a control strategy parameter threshold range;

[0033] The path optimization effect evaluation function is used to perform a global search within the control strategy parameter threshold range to determine the multi-agent collaborative operation control strategy parameters.

[0034] Preferably, the determining of the multi-agent collaborative operation control strategy parameters includes:

[0035] Randomly selecting multiple candidate parameters within the control strategy parameter threshold range, and using the path optimization effect evaluation function to evaluate the effects of the multiple candidate parameters to obtain multiple parameter effect values;

[0036] Performing a search area approximation on the control strategy parameter threshold range based on the multiple parameter effect values ​​to determine a parameter local search area;

[0037] According to the parameter local search area, setting the parameter search step size;

[0038] Parameter search evaluation is performed in the parameter local search area according to the parameter search step, and the search area is iteratively approximated according to the search evaluation result until the preset number of iterations is reached, and the multi-agent collaborative operation control strategy parameters are determined by effect value comparison.

[0039] Preferably, the feedback optimization of the multi-agent collaborative operation control strategy parameters based on the collaborative operation control effect data includes:

[0040] Based on the collaborative operation control effect data, the optimization direction of the multi-agent collaborative operation control strategy parameters is determined to obtain parameter adjustment optimization rules;

[0041] Adjusting and expanding the multi-agent collaborative operation control strategy parameters according to the parameter adjustment optimization rule to obtain a control strategy parameter set;

[0042] Parameter effect comparison and screening are performed within the control strategy parameter set to obtain collaborative control optimization parameters, and multi-agent collaborative operation control is performed using the collaborative control optimization parameters.

[0043] Preferably, the present invention further includes a multi-agent collaborative operation control method based on a wireless deterministic network, which is implemented based on the multi-agent collaborative operation control system based on a wireless deterministic network as described above, and the method includes:

[0044] Collecting initial timestamp data and network transmission delay data of each intelligent agent node in the wireless deterministic network, setting synchronization protocol attributes, performing synchronization calibration processing on the initial timestamp data and network transmission delay data based on the synchronization protocol attributes, and establishing a multi-agent collaborative operation network model;

[0045] By using information analysis technology to associate and obtain task requirement data and agent capability data, a task decomposition method is used to perform instruction conversion processing on the task requirement data and agent capability data to generate a multi-agent collaborative task instruction set;

[0046] Initializing conflict detection parameters based on the multi-agent collaborative operation network model, performing agent path optimization on the conflict detection parameters according to the collaborative operation goal, and iteratively obtaining a multi-agent collaborative optimization path;

[0047] Acquire real-time status data of each agent according to the multi-agent collaborative operation instruction set, perform state correlation analysis based on the real-time status data and the multi-agent collaborative optimization path, and determine the multi-agent collaborative operation control strategy parameters;

[0048] The multi-agent collaborative operation control strategy parameters are simulated and verified to obtain collaborative operation control effect data, and the multi-agent collaborative operation control strategy parameters are feedback optimized based on the collaborative operation control effect data.

[0049] Preferably, the establishment of a multi-agent collaborative operation network model includes:

[0050] Determining network node identification attributes and node communication link attributes based on the synchronization protocol attributes;

[0051] Perform node identification and attribute labeling on the initial timestamp data and the network transmission delay data based on the network node identification attribute to obtain an agent node set and a node attribute set;

[0052] Performing connection relationship analysis on the set of intelligent agent nodes using the initial timestamp data and the network transmission delay data to construct a node adjacency connection relationship set;

[0053] Performing network abstraction on the node adjacency connection relationship set based on the node communication link attributes to obtain a node topology adjacency matrix;

[0054] The agent node set is topologically identified and connected based on the node topology adjacency matrix and the node attribute set to establish the multi-agent collaborative operation network model.

[0055] Preferably, the generating of a multi-agent collaborative operation instruction set includes:

[0056] Use the task decomposition method to perform task splitting and capability matching on the job task requirement data and the agent capability data to obtain a job subtask dataset and an agent adaptation dataset;

[0057] Performing instruction conversion processing on the job subtask data set and the agent adaptation data set respectively through a rule engine to obtain a subtask instruction set and an agent adaptation instruction set;

[0058] Merging and integrating the subtask instruction set and the agent adaptation instruction set to generate an initial collaborative operation instruction set;

[0059] An instruction verifier is used to verify the compliance of the initial collaborative operation instruction set to obtain the multi-agent collaborative operation instruction set.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] The time synchronization calibration module achieves high-precision time synchronization among agent nodes in a wireless deterministic network. This module collects initial timestamp data and network transmission delay data, performs calibration based on synchronization protocol properties, and establishes a multi-agent collaborative operation network model. By determining network node identification attributes and communication link properties and constructing a node topology adjacency matrix, each agent node maintains high consistency in the time dimension, significantly reducing action execution deviations caused by time asynchrony. This establishes a precise time benchmark for multi-agent collaborative operations and improves the system's synchronization control accuracy.

[0062] The task instruction distribution module uses information parsing technology to correlate task requirements with agent capability data, and combines this with task decomposition methods to generate a dynamically adaptable set of collaborative task instructions. This module utilizes a rules engine for instruction conversion and a command verifier to ensure instruction compliance, achieving a precise match between tasks and agent capabilities. Compared to traditional fixed task allocation strategies, this module dynamically adjusts task allocation based on the real-time capabilities of agents, improving task execution efficiency and resource utilization. This is particularly suitable for scenarios with a large number of agents and complex tasks.

[0063] The path conflict detection module initializes conflict detection parameters and optimizes paths based on the collaborative network model. It iteratively determines the optimal collaborative path by constructing a path optimization evaluation function. This module can detect potential conflicts in agent path planning in real time and dynamically adjust path parameters to avoid collisions. This reduces interruptions caused by path conflicts and improves the safety and smoothness of multi-agent movement in complex topological networks. Furthermore, an iterative optimization mechanism ensures the global optimality of path planning, further enhancing operational efficiency.

[0064] The collaborative state perception module and the control parameter optimization module form a closed-loop feedback mechanism, enabling dynamic control of the entire collaborative operation process. The collaborative state perception module determines control strategy parameters by analyzing the correlation between real-time state data and the optimization path. The control parameter optimization module continuously adjusts parameters to adapt to changes in the operating environment through simulation verification and feedback optimization. This closed-loop mechanism enables the system to respond to changes in agent state and network dynamics in real time, significantly improving the robustness and adaptability of the control system and ensuring efficient collaborative operation even in emergencies such as agent failures and network latency fluctuations.

[0065] This invention improves the overall stability, efficiency, and reliability of multi-agent collaborative operations, making it suitable for a variety of complex scenarios, including intelligent warehousing, industrial manufacturing, and drone collaboration. By combining the characteristics of wireless deterministic networks with multi-module collaborative design, it overcomes the limitations of traditional wireless communication environments on multi-agent collaboration, providing a feasible technical solution for the practical application of large-scale multi-agent collaborative operations, with high practical value and promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a working principle diagram of the multi-agent collaborative operation control system based on wireless deterministic network according to the present invention;

[0067] Figure 2 A flowchart for establishing a multi-agent collaborative operation network model;

[0068] Figure 3 A flowchart for generating a multi-agent collaborative operation instruction set;

[0069] Figure 4 Flowchart for iteratively obtaining multi-agent collaborative optimization paths;

[0070] Figure 5 Flowchart for determining control strategy parameters for multi-agent collaborative operations. DETAILED DESCRIPTION

[0071] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0072] See also Figure 1-Figure 5 The present invention provides a multi-agent collaborative task control method and system based on a wireless deterministic network. The system includes: a time synchronization calibration module, a task instruction distribution module, a path conflict detection module, a collaborative state perception module, and a control parameter optimization module. The specific implementation steps are as follows:

[0073] The time synchronization calibration module collects the initial timestamp data and network transmission delay data of each intelligent node in the wireless deterministic network, sets the synchronization protocol attributes, performs synchronization calibration processing on the initial timestamp data and network transmission delay data based on the attributes, and establishes a multi-agent collaborative operation network model.

[0074] The task instruction distribution module obtains the operation task requirement data and agent capability data through information analysis technology, uses the task decomposition method to perform instruction conversion processing on the operation task requirement data and agent capability data, and generates a multi-agent collaborative operation instruction set.

[0075] The path conflict detection module initializes the conflict detection parameters based on the multi-agent collaborative operation network model, optimizes the agent path of the conflict detection parameters according to the collaborative operation goals, and iteratively obtains the multi-agent collaborative optimization path.

[0076] The collaborative state perception module obtains the real-time state data of each agent according to the multi-agent collaborative operation instruction set, performs state correlation analysis based on the real-time state data and the multi-agent collaborative optimization path, and determines the multi-agent collaborative operation control strategy parameters.

[0077] The control parameter optimization module simulates and verifies the control strategy parameters of multi-agent collaborative operations, obtains collaborative operation control effect data, and performs feedback optimization on the control strategy parameters of multi-agent collaborative operations based on the collaborative operation control effect data.

[0078] Example 1: The time synchronization calibration module in the system establishes a multi-agent collaborative operation network model. The specific implementation method is as follows:

[0079] Collect initial timestamp data from each intelligent node in the wireless deterministic network. This data represents the initial time each intelligent node enters the network and contains its own time information. Also collected is network transmission delay data, which reflects the time delay experienced during data transmission within the network and is related to factors such as the network's transmission medium and the distance between nodes. Next, set the synchronization protocol properties. These properties define the rules and methods for time synchronization within the network, such as the synchronization period and accuracy requirements.

[0080] Based on the set synchronization protocol properties, the collected initial timestamp data and network transmission delay data are synchronized and calibrated. During this process, the network node identification attributes and node communication link attributes are determined based on the synchronization protocol properties. The network node identification attributes uniquely identify each intelligent node in the network and contain information such as the node number and type. The node communication link attributes describe the characteristics of the communication link between nodes, such as the link bandwidth and delay characteristics.

[0081] Node identification and attribute annotation are performed on the initial timestamp data and network transmission delay data based on the network node identification attributes. For the initial timestamp data, the network node identification attributes are used to determine the agent node corresponding to each timestamp, and the relevant attributes of the node are annotated. For the network transmission delay data, the sending and receiving nodes corresponding to the delay data are also determined based on the network node identification attributes, and the attributes of the link are annotated. Through this process, an agent node set and a node attribute set are obtained. The agent node set contains the identification information of all agent nodes in the network, and the node attribute set records the detailed attributes of each node.

[0082] The connectivity of the agent node set is analyzed using initial timestamp data and network transmission delay data. Initial timestamp data reflects the interaction between nodes for time synchronization, while network transmission delay data reveals the path and delay of data transmission between nodes. By analyzing this data, the connectivity between nodes is determined, and a node adjacency connection set is constructed. This node adjacency connection set records which nodes have direct communication connections.

[0083] A network abstraction is performed on the set of node adjacency connections based on the properties of node communication links. Properties such as bandwidth and latency in these properties are used to abstract the actual node adjacency connections into a mathematical representation, thereby obtaining a node topology adjacency matrix. The node topology adjacency matrix is ​​a mathematical matrix whose elements represent the connection status and characteristics between nodes.

[0084] The agent node set is topologically identified and connected based on the node topology adjacency matrix and the node attribute set. The node topology adjacency matrix defines the topological structure between nodes, while the node attribute set provides detailed attribute information for each node. Combining these two, each agent node is accurately identified and connected within the topology, thereby establishing a multi-agent collaborative operation network model. This model clearly describes the distribution of agent nodes in the network, the connections between nodes, and the node attributes, providing a foundational network structure model for subsequent multi-agent collaborative operation control. Throughout the entire establishment process, each step is closely dependent on the results of the previous step. Through the gradual processing and abstraction of data, an accurate network model is ultimately formed. For example, during node identification and attribute labeling, accurate identification and labeling are crucial for the subsequent construction of the connection relationship set. Correctly constructing the node adjacency connection relationship set is the foundation for network abstraction and the topological adjacency matrix. Only when each step is completed accurately can the established multi-agent collaborative operation network model conform to the actual network conditions and provide reliable support for subsequent collaborative operation control.

[0085] Example 2: The implementation method of the task instruction distribution module generating a multi-agent collaborative operation instruction set is as follows:

[0086] Through information parsing technology, we correlate and obtain job task requirement data and agent capability data. Job task requirement data includes information such as the specific goals, workload, and completion time requirements of the job. For example, a job may require a specific number of tasks to be completed within a specific timeframe. Agent capability data includes information such as the hardware performance, software functionality, and communication capabilities of each agent. For example, a particular agent may possess specific sensor detection capabilities or data processing speeds.

[0087] The task decomposition method is used to process the job task requirement data and agent capability data. The task decomposition method needs to consider factors such as the nature of the task and the capability characteristics of the agent, and split the overall job task into multiple subtasks. During the task splitting process, capability matching is performed at the same time, that is, analyzing what kind of agent capabilities are required to complete each subtask, and matching each subtask in the job task requirement data with an agent with corresponding capabilities in the agent capability data. For example, if there is a data acquisition subtask that requires an agent to have the ability to collect signals in a specific frequency band, an agent with this capability is found from the agent capability data for matching, thereby obtaining a job subtask dataset and an agent adaptation dataset. The job subtask dataset contains detailed information on each subtask after splitting, and the agent adaptation dataset records the adaptation agent information corresponding to each subtask.

[0088] The rule engine performs instruction conversion processing on the job subtask dataset and the agent adaptation dataset respectively. The rule engine has a series of conversion rules preset, which are formulated based on the type of job task, the interface protocol of the agent, etc. For the job subtask dataset, the rule engine converts it into subtask instructions that the agent can understand and execute according to the requirements and goals of each subtask. For example, the data collection subtask is converted into instructions containing specific parameters such as collection frequency and collection range; for the agent adaptation dataset, the rule engine converts the adaptation information into agent adaptation instructions based on the communication protocol and interface specification of the agent, such as specifying the communication address and data transmission format for a certain agent to execute the subtask, thereby obtaining the subtask instruction set and the agent adaptation instruction set.

[0089] Merge and integrate the subtask instruction set and the agent adaptation instruction set. When merging, it is necessary to ensure that each subtask instruction is accurately associated with the corresponding agent adaptation instruction, that is, to clarify which agent executes which subtask instruction. For example, merge the data collection subtask instruction with the corresponding adaptation instruction of the agent with collection capabilities to form a complete collaborative operation instruction, which contains information such as task content and execution subject, thereby generating the initial collaborative operation instruction set. The initial collaborative operation instruction set contains the instruction information of all subtasks and their corresponding execution agents, but there may be problems such as the instruction format not meeting the requirements or the instruction content conflicting.

[0090] The initial collaborative work instruction set is verified for compliance using an instruction verifier. The instruction verifier checks each instruction in the initial collaborative work instruction set against pre-set compliance standards, including whether the instruction format is correct, the parameters are complete, and whether there are any conflicts. For example, it checks whether the time parameters in the instructions are within a reasonable range and whether the agent's address is correct. Instructions that pass verification are retained in the instruction set; instructions that fail verification are marked and returned for revision. After verification by the instruction verifier, the multi-agent collaborative work instruction set is obtained. This instruction set has been rigorously processed and verified to ensure that the multi-agent collaborative work instruction set accurately understands and executes the corresponding task instructions, achieving the goal of collaborative work. Every step in the generation process is essential. Accurate data acquisition through information parsing technology is the foundation, task decomposition and capability matching are key, rule engine conversion ensures the executable nature of the instructions, merging and integration form a system, and compliance verification ensures the correctness of the instructions. Each step requires meticulous processing to ensure that the final multi-agent collaborative work instruction set meets the actual operational needs and the execution requirements of the agents. For example, during task decomposition and capability matching, improperly split subtasks or inaccurate capability matching can lead to problems with subsequent instruction conversion and execution. Failure to detect errors in instructions during instruction verification can lead to misunderstandings or even inability to execute tasks. Therefore, each step must be strictly carried out according to prescribed methods and processes to ensure the accuracy and reliability of the generated instruction set.

[0091] Example 3: The implementation method of iteratively obtaining a multi-agent collaborative optimization path by the path conflict detection module is as follows:

[0092] Conflict assessment indicators are extracted for collaborative work objectives. Collaborative work objectives may involve multiple aspects, such as task completion time, agent energy consumption, and task execution accuracy. Indicators that can reflect the degree of path conflict need to be extracted from these objectives. For example, if the collaborative work goal is to complete the task in the shortest time, then time delays on the path and the possibility of collisions between agents may become important conflict assessment indicators. By analyzing and decomposing the collaborative work goals, a set of conflict assessment indicators is obtained. This set contains multiple specific indicators for evaluating path conflicts. Then, based on the set of conflict assessment indicators, a path optimization effect evaluation function is constructed. The path optimization effect evaluation function is a mathematical expression that integrates various conflict assessment indicators to quantitatively evaluate the optimization effect of the path. The construction of the function needs to consider the weight of each indicator and the relationship between them. For example, different weights can be assigned to each indicator based on its importance to the collaborative work goal.

[0093] The position and velocity of the detection particle are determined based on the conflict detection parameters. These parameters include the agent's initial position, target position, and speed limit, among other parameters related to the agent's movement. A detection particle can be understood as an abstract concept representing the agent's possible paths. The position of each detection particle corresponds to a possible position on the path, while the velocity represents the particle's movement rate in the solution space. For example, for an agent with an initial position of (1,1) and a target position of (5,5), and a speed limit of no more than 2 units per second in the conflict detection parameters, the initial position of the detection particle may be near the initial position, while the velocity is set according to the speed limit. The effectiveness of the conflict detection parameters is evaluated using a constructed path optimization evaluation function. Substituting the conflict detection parameters into the evaluation function, the effect value corresponding to each detection particle is calculated, resulting in a detection particle effect set. This set records the optimization effect of each detection particle under the current parameters.

[0094] The position and velocity of the detection particles are iteratively updated based on the set of detection particle effects. This iterative update process follows specific algorithmic rules, such as the update rules used in particle swarm optimization. In each iteration, the position and velocity of the detection particle are adjusted based on its current effect value. If a detection particle has a good effect value, indicating that the path it represents is more optimal, its velocity is adjusted according to specific rules to move it in the more optimal direction, and its position is updated simultaneously. For detection particles with poor effect values, their velocity and position are also adjusted according to the corresponding rules to explore new possible paths. This iterative process continues until a preset stopping condition is met. The preset stopping condition can be reaching a certain number of iterations or when the effect value of the detection particle changes very little over multiple consecutive iterations, indicating that the solution is close to the optimal solution.

[0095] When the preset stopping condition is met, the particle with the best effect is selected from the set of detection particle effects. The position corresponding to this optimal detection particle is the agent's optimized path position, representing the agent's optimal path position under the current conflict detection parameters and collaborative operation goals. For example, after multiple iterations, if a detection particle has the highest effect value in the set, it indicates that the path it represents best meets the collaborative operation goals and reduces path conflicts.

[0096] Based on the optimized positions of agent paths, the multi-agent collaborative network model is updated and adjusted. The multi-agent collaborative network model records the positions and connections of agent nodes. Based on the optimized positions of the agent paths, the agent path information in the model is updated, and the agent movement paths and node connections in the network are redefined, thereby obtaining the optimal multi-agent collaborative path. This optimized path effectively reduces path conflicts between agents and improves the efficiency and reliability of collaborative work. Throughout the iterative process, each iteration is an attempt to optimize the path. By continuously adjusting the positions and velocities of the detection particles, the optimal path is gradually approached. Accurately extracting conflict assessment indicators and constructing a reasonable evaluation function for path optimization are key to ensuring the effectiveness of the iterative process. These indicators determine the accuracy of the evaluation of path optimization, thus affecting the update direction of the detection particles and the quality of the resulting optimized path. For example, if the conflict assessment indicator fails to fully reflect the actual path conflicts, the constructed evaluation function will be biased, resulting in the inability to find the true optimal path during the iterative process. Furthermore, if the iterative update rules are inappropriate, the iterative process may converge slowly or become trapped in a local optimum, failing to obtain the global optimal path. Therefore, each link needs to be carefully designed and strictly implemented to ensure that the final multi-agent collaborative optimization path can meet the actual needs of collaborative work.

[0097] Example 4: When the collaborative state perception module determines the parameters of the multi-agent collaborative operation control strategy, it needs to be processed in detail in combination with the specific scenario. Taking the multi-robot collaborative sorting operation in the logistics warehouse as an example, the network traffic distribution impact analysis is performed based on the multi-agent collaborative optimization path. Assuming that the multi-agent collaborative optimization path plans the movement route from the sorting starting point to different shelves and shipping ports for each robot, it is necessary to analyze the network traffic distribution on these paths. For example, when multiple robots move along a certain channel at the same time, the network communication traffic of the channel will increase, which may cause data transmission delays. By analyzing the traffic changes on this path, the network traffic distribution influencing parameters are obtained, where the path load influencing parameters can be expressed as the number of robots or data transmission volume on each path, and the collaborative operation influencing parameters may involve operation time delays caused by path congestion.

[0098] A collaborative control strategy library is constructed, storing various control strategies tailored to different traffic distributions and operational states. For example, when a path is highly loaded, a strategy might adjust the robot's speed or re-plan its path. When collaborative work is impacted, a strategy might adjust the order of task assignments. Based on each agent's real-time status data and network traffic distribution influencing parameters, a matching and selection process is performed against the collaborative control strategy library. Real-time status data includes the robot's current location, task completion progress, and battery level. For example, a robot is currently located near shelf A, has completed 30% of its sorting task, and has 60% battery remaining. This data is combined with network traffic distribution influencing parameters (e.g., a path currently has five robots moving simultaneously, resulting in a high load). A matching strategy is then searched for in the strategy library. Assuming a strategy is found that "when the load on a path exceeds a threshold, instruct subsequent robots to slow down and re-plan a temporary path," this represents a multi-agent collaborative matching control strategy.

[0099] Based on a multi-agent collaborative matching control strategy, threshold ranges are assigned to parameters influencing real-time status data and network traffic distribution. For example, for the path load influencing parameter, threshold ranges are set for light load, medium load, and heavy load: when there are fewer than three robots on the path, it is considered light load; when there are 3-5 robots, it is considered medium load; and when there are more than five robots, it is considered heavy load. For the collaborative operation influencing parameter, threshold ranges are set for different degrees of delay. By dividing these threshold ranges, threshold ranges for control strategy parameters are derived, such as the speed adjustment range under heavy load and the time interval range for path replanning.

[0100] A global search is performed within the control strategy parameter threshold range using a path optimization evaluation function. This function comprehensively considers factors such as operational efficiency and the degree of path conflict. For example, the total time to complete the sorting task and the probability of robot collision risk are used as evaluation indicators. Within the threshold range, different control strategy parameter combinations are searched and evaluated. For example, within the heavy load threshold range, different speed adjustment ranges (e.g., 20%, 30%, and 40% deceleration) and different path replanning intervals (e.g., every 10 seconds, 15 seconds, and 20 seconds) are tried. These parameter combinations are then substituted into the evaluation function to calculate the effect values. By comparing the effect values ​​of different parameter combinations, the control strategy parameters for multi-agent collaborative operation are determined. For example, after a search and evaluation, the evaluation function's effect value is found to be optimal when the speed is reduced by 30% and the path is replanned every 15 seconds. These parameters are then used as the control strategy parameters.

[0101] Throughout the entire process, using a logistics warehouse sorting scenario as an example, each step is closely linked. When analyzing network traffic distribution, the load situation must be calculated in real time based on the actual robot movement path and communication data. When matching control strategies, the appropriate strategy must be selected from the strategy library based on the robot's real-time status and traffic parameters. The threshold range must be set to a reasonable interval based on operational experience and system performance. A global search is performed by repeatedly trying different parameter combinations to find the optimal solution. For example, if the load on a certain path in a warehouse area reaches the heavy load threshold due to the simultaneous passage of multiple robots, the collaborative state perception module will match a path replanning strategy from the strategy library based on the robot's real-time position, task progress, and other data. Then, within the set speed adjustment and planning time interval thresholds, an evaluation function search is used to determine the specific deceleration ratio and planning interval, enabling the robot to alleviate path congestion while minimizing the impact on sorting efficiency. This ultimately determines the control strategy parameters and ensures the smooth operation of multi-agent collaborative operations.

[0102] Example 5: The control parameter optimization module performs feedback optimization on the control strategy parameters of multi-agent collaborative operations based on collaborative operation control performance data. This detailed implementation is described using a multi-UAV inspection operation scenario in a smart grid as an example. In this scenario, multiple UAVs must perform transmission line inspections according to established control strategy parameters, including flight speed, data transmission frequency, and inspection path switching thresholds.

[0103] Based on the collaborative operation control effect data, the optimization direction of the multi-agent collaborative operation control strategy parameters is determined. The collaborative operation control effect data comes from the actual execution of the drone inspection operation. For example, during the inspection process, a drone's image acquisition is blurred due to its high flight speed, affecting the quality of the inspection data; or multiple drones have unreasonable data transmission frequency settings, causing network channel congestion, resulting in delayed upload of inspection data. By analyzing these effect data, the problems with the current control strategy parameters are determined, and then the parameter adjustment optimization rules are obtained. For example, if it is found that the image quality has deteriorated due to excessive flight speed, the optimization rule may be "when the image clarity is lower than the preset standard, reduce the flight speed parameter"; if the data transmission delay exceeds the threshold, the optimization rule may be "increase the data transmission interval and reduce the transmission frequency parameter."

[0104] Based on parameter adjustment optimization rules, the control strategy parameters for multi-agent collaborative operations are adjusted and expanded to obtain a control strategy parameter set. For example, the flight speed parameter, originally set in the control strategy for a flight speed of 20 m / s, needs to be reduced according to the optimization rules. In this case, the original parameter is expanded to generate multiple candidate parameters, such as 18 m / s, 16 m / s, and 15 m / s. For the data transmission frequency parameter, originally set at 5 data transmissions per second, if the optimization rules require a lower transmission frequency, the expanded candidate parameters might be 4, 3, or 2 data transmissions per second. These adjusted candidate parameters are combined to form a control strategy parameter set, which contains multiple possible parameter combinations, such as (flight speed 18 m / s, data transmission frequency 4 times / s) and (flight speed 16 m / s, data transmission frequency 3 times / s).

[0105] Parameter effects are compared and screened within the control strategy parameter set to obtain optimized collaborative control parameters. These parameters are then used to implement multi-agent collaborative control. Each parameter combination is then validated in a simulated inspection scenario. For example, for a parameter combination of (flight speed 18 m / s, data transmission frequency 4 times / s), a drone inspection process is simulated under these parameters to observe whether image clarity improves and whether network transmission congestion persists. Similarly, for a parameter combination of (flight speed 16 m / s, data transmission frequency 3 times / s), a simulation is performed, recording image clarity and data transmission delay.

[0106] By comparing the performance of different parameter combinations in simulations, the optimal combination is selected. Assume that, in the simulation, using the parameter combination (flight speed of 16 meters per second, data transmission frequency of 3 times per second), the image clarity collected by the drone meets the standard requirements and the data transmission delay is reduced to an acceptable range. However, other parameter combinations either still fail to meet the image clarity requirements or fail to effectively improve data transmission delay. This parameter combination is then determined to be the optimized parameter for collaborative control.

[0107] The collaborative control optimization parameters are applied to actual multi-UAV inspection operations. For example, new control instructions are sent to all participating UAVs, adjusting their flight speed to 16 meters per second and their data transmission frequency to three times per second. During subsequent inspections, the collaborative control performance data is continuously monitored. If image quality or data transmission issues are detected again, the optimization process is repeated, and the control strategy parameters are further optimized to adapt to different operating environments and changing requirements.

[0108] Throughout the feedback optimization process, using smart grid drone inspections as an example, each step is closely centered around practical operational issues. Optimization direction determination is based on real-world operational performance data, ensuring targeted adjustment targets. Parameter adjustment and expansion are based on sound optimization rules to avoid blind parameter modifications. Parameter performance comparison and screening are conducted through simulation verification to ensure the selected parameter combinations are effective in real-world applications. For example, if weather conditions cause a drone's flight stability to decrease during an inspection, the original flight speed parameters may cause increased wobbling. Analysis of the performance data reveals that image blurring is related to flight speed and weather factors. Optimization rules are then developed. In addition to adjusting the flight speed parameter, optimization of the flight attitude adjustment parameter may also be required. This expands the set of parameter combinations encompassing both flight speed and attitude adjustments. Through comparison and screening, coordinated control optimization parameters suitable for inclement weather are identified, enabling the drone to efficiently complete inspection missions even in complex environments, achieving dynamic optimization and feedback adjustment of control strategy parameters.

[0109] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0110] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-agent collaborative operation control system based on a wireless deterministic network, characterized by: The system comprises: A time synchronization calibration module, which is used to collect initial timestamp data and network transmission delay data of each intelligent node in the wireless deterministic network, set synchronization protocol attributes, perform synchronization calibration processing on the initial timestamp data and network transmission delay data based on the synchronization protocol attributes, and establish a multi-agent collaborative operation network model; A task instruction distribution module is used to obtain task requirement data and agent capability data through information analysis technology, perform instruction conversion processing on the task requirement data and agent capability data using a task decomposition method, and generate a multi-agent collaborative task instruction set; A path conflict detection module, the path conflict detection module is used to initialize conflict detection parameters based on the multi-agent collaborative operation network model, perform agent path optimization on the conflict detection parameters according to the collaborative operation goal, and iteratively obtain the multi-agent collaborative optimization path; A collaborative state perception module, which is used to obtain real-time state data of each agent according to the multi-agent collaborative operation instruction set, perform state correlation analysis based on the real-time state data and the multi-agent collaborative optimization path, and determine the multi-agent collaborative operation control strategy parameters; A control parameter optimization module, which is used to simulate and verify the control strategy parameters of the multi-agent collaborative operation, obtain collaborative operation control effect data, and perform feedback optimization on the control strategy parameters of the multi-agent collaborative operation based on the collaborative operation control effect data; The iterative process of obtaining a multi-agent collaborative optimization path includes: Extracting conflict assessment indicators for the collaborative operation target to obtain a conflict assessment indicator set, and constructing a path optimization effect evaluation function based on the conflict assessment indicator set; Determine the detection particle position and the detection particle speed according to the conflict detection parameters, and use the path optimization effect evaluation function to evaluate the effect of the conflict detection parameters to obtain a detection particle effect set; Iteratively updating the detection particle position and the detection particle velocity based on the detection particle effect set until a preset stop condition is met, and screening and determining the detection particle with the best effect, wherein the best detection particle includes the intelligent agent path optimization position; The detection particle represents the abstract concept of the agent's path, and the position of each detection particle corresponds to a position of the agent on the path; Based on the agent path optimization position, the multi-agent collaborative operation network model is updated and adjusted to obtain the multi-agent collaborative optimization path; The determining of the multi-agent collaborative operation control strategy parameters includes: Performing a network traffic distribution impact analysis based on the multi-agent collaborative optimization path to obtain network traffic distribution impact parameters, wherein the network traffic distribution impact parameters include path load impact parameters and collaborative operation impact parameters; The collaborative operation influencing parameters include operation time delay caused by path congestion; Constructing a collaborative control strategy library, and performing matching screening with the collaborative control strategy library based on the real-time status data and the network traffic distribution influencing parameters to obtain a multi-agent collaborative matching control strategy; Based on the multi-agent collaborative matching control strategy, the real-time status data and the network traffic distribution influencing parameters are divided into threshold ranges to obtain a control strategy parameter threshold range; The threshold range division includes: setting light load, medium load, and heavy load threshold ranges for the path load influencing parameter; setting different delay threshold ranges for the operation time delay in the collaborative operation influencing parameter; The control strategy parameter threshold range includes the speed adjustment amplitude range under heavy load and the time interval range for path replanning; Performing a global search within the control strategy parameter threshold range using the path optimization effect evaluation function to determine the multi-agent collaborative operation control strategy parameters; The feedback optimization of the multi-agent collaborative operation control strategy parameters based on the collaborative operation control effect data includes: Based on the collaborative operation control effect data, the optimization direction of the multi-agent collaborative operation control strategy parameters is determined to obtain parameter adjustment optimization rules; Adjusting and expanding the multi-agent collaborative operation control strategy parameters according to the parameter adjustment optimization rule to obtain a control strategy parameter set; Parameter effect comparison and screening is performed within the control strategy parameter set, and the parameter combination with the best effect is screened out by comparing the performance of different parameter combinations in simulation verification; the parameter combination with the best effect is the collaborative control optimization parameter, and multi-agent collaborative operation control is performed through the collaborative control optimization parameter.

2. The multi-agent collaborative operation control system based on a wireless deterministic network according to claim 1, characterized in that: The establishment of a multi-agent collaborative operation network model includes: Determining network node identification attributes and node communication link attributes based on the synchronization protocol attributes; Perform node identification and attribute labeling on the initial timestamp data and the network transmission delay data based on the network node identification attribute to obtain an agent node set and a node attribute set; Performing connection relationship analysis on the set of intelligent agent nodes using the initial timestamp data and the network transmission delay data to construct a node adjacency connection relationship set; Performing network abstraction on the node adjacency connection relationship set based on the node communication link attributes to obtain a node topology adjacency matrix; The agent node set is topologically identified and connected based on the node topology adjacency matrix and the node attribute set to establish the multi-agent collaborative operation network model.

3. The multi-agent collaborative operation control system based on a wireless deterministic network as claimed in claim 1, characterized in that: The generating of a multi-agent collaborative operation instruction set includes: Use the task decomposition method to perform task splitting and capability matching on the job task requirement data and the agent capability data to obtain a job subtask dataset and an agent adaptation dataset; Performing instruction conversion processing on the job subtask data set and the agent adaptation data set respectively through a rule engine to obtain a subtask instruction set and an agent adaptation instruction set; Merging and integrating the subtask instruction set and the agent adaptation instruction set to generate an initial collaborative operation instruction set; An instruction verifier is used to verify the compliance of the initial collaborative operation instruction set to obtain the multi-agent collaborative operation instruction set.

4. The multi-agent collaborative operation control system based on a wireless deterministic network according to claim 1, characterized in that: The determining of the multi-agent collaborative operation control strategy parameters includes: Randomly selecting multiple candidate parameters within the control strategy parameter threshold range, and using the path optimization effect evaluation function to evaluate the effects of the multiple candidate parameters to obtain multiple parameter effect values; Performing a search area approximation on the control strategy parameter threshold range based on the multiple parameter effect values ​​to determine a parameter local search area; According to the parameter local search area, setting the parameter search step size; Parameter search evaluation is performed in the parameter local search area according to the parameter search step, and the search area is iteratively approximated according to the search evaluation result until the preset number of iterations is reached, and the multi-agent collaborative operation control strategy parameters are determined by effect value comparison.

5. A multi-agent collaborative operation control method based on a wireless deterministic network, characterized in that: Implementation of a multi-agent collaborative operation control system based on a wireless deterministic network according to any one of claims 1 to 4, the method comprising: Collecting initial timestamp data and network transmission delay data of each intelligent agent node in the wireless deterministic network, setting synchronization protocol attributes, performing synchronization calibration processing on the initial timestamp data and network transmission delay data based on the synchronization protocol attributes, and establishing a multi-agent collaborative operation network model; By using information analysis technology to associate and obtain task requirement data and agent capability data, a task decomposition method is used to perform instruction conversion processing on the task requirement data and agent capability data to generate a multi-agent collaborative task instruction set; Initializing conflict detection parameters based on the multi-agent collaborative operation network model, performing agent path optimization on the conflict detection parameters according to the collaborative operation goal, and iteratively obtaining a multi-agent collaborative optimization path; Acquire real-time status data of each agent according to the multi-agent collaborative operation instruction set, perform state correlation analysis based on the real-time status data and the multi-agent collaborative optimization path, and determine the multi-agent collaborative operation control strategy parameters; The multi-agent collaborative operation control strategy parameters are simulated and verified to obtain collaborative operation control effect data, and the multi-agent collaborative operation control strategy parameters are feedback optimized based on the collaborative operation control effect data.

6. The multi-agent collaborative operation control method based on a wireless deterministic network according to claim 5, characterized in that: The establishment of a multi-agent collaborative operation network model includes: Determining network node identification attributes and node communication link attributes based on the synchronization protocol attributes; Perform node identification and attribute labeling on the initial timestamp data and the network transmission delay data based on the network node identification attribute to obtain an agent node set and a node attribute set; Performing connection relationship analysis on the set of intelligent agent nodes using the initial timestamp data and the network transmission delay data to construct a node adjacency connection relationship set; Performing network abstraction on the node adjacency connection relationship set based on the node communication link attributes to obtain a node topology adjacency matrix; The agent node set is topologically identified and connected based on the node topology adjacency matrix and the node attribute set to establish the multi-agent collaborative operation network model.

7. The multi-agent collaborative task control method based on a wireless deterministic network according to claim 5, characterized in that: The generating of a multi-agent collaborative operation instruction set includes: Use the task decomposition method to perform task splitting and capability matching on the job task requirement data and the agent capability data to obtain a job subtask dataset and an agent adaptation dataset; Performing instruction conversion processing on the job subtask data set and the agent adaptation data set respectively through a rule engine to obtain a subtask instruction set and an agent adaptation instruction set; Merging and integrating the subtask instruction set and the agent adaptation instruction set to generate an initial collaborative operation instruction set; An instruction verifier is used to verify the compliance of the initial collaborative operation instruction set to obtain the multi-agent collaborative operation instruction set.

Images