Multi-agent cooperative control system and method

Abstract

The embodiment of the invention discloses a multi-agent cooperative control system and method, the multi-agent control system specifically divides system network nodes, specifically sets node control functions of agents in the system, does not need a complex cooperative control algorithm, considers the connectivity of the network nodes under complex conditions, and also considers the connectivity of the network nodes under complex conditions. And the calculation consumption burden reduction caused by using a complex algorithm is also considered, the overall collaboration is completed by using a relatively convenient method, and the stability and the efficiency of the topological structure of the system are kept as much as possible in the same target task process.

Description

Technical Field

[0001] The present invention relates to the field of multi-agent collaborative control, and more particularly to a multi-agent collaborative control system and method. Background Art

[0002] With the wide application of intelligent systems in fields such as industry, logistics, security, and transportation, collaborative control technology based on multi-agent (Multi-Agent System, MAS) has received extensive attention. The combination of complex network structures and intelligent agent nodes enables the system to have stronger adaptability and self-organization capabilities. Existing research has proposed various theoretical solutions such as synchronous control algorithms and dynamic topology adaptation mechanisms, but there are obvious defects in actual deployment, such as the connectivity and stability problems of multi-agent nodes in the system.

[0003] Most existing systems assume that the network is always connected or jointly connected, and derive collaborative rules under ideal topologies. However, in actual scenarios, due to factors such as limited communication capabilities of agents, node hardware failures, large variations in task loads, and insufficient universality of node control functions, the connectivity of the system is vulnerable to attacks and disruptions, resulting in collaborative failures. In addition, some solutions adopt forced adjacency maintenance strategies to maintain connectivity. Although local connections are guaranteed, the mobility and flexible control capabilities of the system are severely affected, contrary to the original intention of efficient collaborative control. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to at least partially consider the above problems. And a multi-agent collaborative control and method is proposed for some of these problems. More specifically, please refer to the following:

[0005] In one aspect of the present invention, a multi-agent collaborative control system is provided, including a plurality of intelligent agent nodes, characterized in that:

[0006] Each of the intelligent agent nodes at least has a node intelligent processing unit, a state perception and collection unit, a task execution and control unit, and a function monitoring and adjustment unit;

[0007] Each intelligent agent node of the collaborative control system is classified into node categories, specifically classified as: a master control node, a collaborative control node, and an execution control node. All nodes and the connection relationships of the nodes form a collaborative control network;

[0008] The collaborative control system determines the collaborative control law of the nodes based on a composite potential function.

[0009] Preferably, the function monitoring and adjustment unit is used to monitor the energy consumption of the current intelligent agent or adjust the radiation characteristics of the intelligent agent, and adjust the radiation distribution of the intelligent agent according to the real-time system working requirements.

[0010] Preferably, the selection of the master control node is determined based on the radiation power adjustment performance, energy consumption, integrity of the NIPU unit, and node location characteristics of the current agent node.

[0011] Preferably, the cooperative control node is selected and constructed based on a clustering algorithm with multi-criterion priority sorting for cooperative control.

[0012] Preferably, this algorithm determines the priority function by comprehensively evaluating the spatial position, node degree, communication radiation ability, and unique identity identifier of each agent, as shown below:

[0013] Z(m)=Z sp (m)⊕Z nd (m)⊕Z fq (m)⊕Z vp (m)⊕Z id (m);

[0014] Among them, ⊕ represents the bit concatenation operation logic. After converting each quantization index into a fixed-length binary code, they are concatenated in high-order priority according to the set order to construct a new fixed-length bit string, which is ultimately used for node priority comparison and sorting.

[0015] Preferably, after removing the master control node and the cooperative control node from the cooperative control system, the remaining ones are the execution control nodes.

[0016] Preferably, the cooperative control system determines the cooperative control law of the agent node based on a composite artificial potential function.

[0017] Preferably, if a lost node appears during the obstacle avoidance event of the cooperative control system, the master control node adjusts the current agent radiation characteristics to achieve the connection restoration of the lost node.

[0018] Another aspect of the present invention proposes a control method for a multi-agent cooperative control system, which is characterized by including the following steps:

[0019] First step, the system is initialized and started. All agent nodes complete initialization according to the preset parameters carried by themselves, including communication parameter setting, attitude sensor calibration, and communication neighborhood radius setting;

[0020] Second step, the cooperative control network is constructed, and all agent nodes in the system are divided according to the set requirements;

[0021] Third step, based on the composite potential function, the control law of the node agent is constructed. The composite potential function includes at least a cooperative connectivity maintenance potential function U1, an execution control attraction potential function U2, a target attraction potential function U3, and an obstacle avoidance repulsion potential function U4;

[0022] Step 4, dynamic monitoring response of the collaborative control system. When a node loses connection or is out of the communication range after being attacked during the obstacle avoidance process of the collaborative control system, the master control node automatically triggers the radiation performance adjustment function, quickly recovers through hardware adjustment, and updates the collaborative control network structure in a timely manner;

[0023] Step 5, periodic update of the collaborative control system. The above steps are cyclically executed within a preset control period to form an adaptive closed-loop control.

[0024] Preferably, for the construction of the collaborative control network, the master control node is selected according to the radiation performance, energy consumption, integrity of the NIPU unit, and position weight of the agent node; the collaborative control nodes are selected according to the multi-index bit splicing mechanism; the rest are execution control nodes; the streamlined backbone subnet of the system is extracted by the minimum spanning tree method, and the redundant collaborative network connections are trimmed to generate the collaborative control network.

[0025] The present invention discloses a multi-agent collaborative control system and a collaborative control method, which do not require complex collaborative control algorithms. It not only considers the connectivity of network nodes in complex situations but also takes into account the reduction of the computational consumption burden brought by using complex algorithms. The overall collaboration is completed by a relatively convenient method. The priority function structure during the control process can be configured or fine-tuned according to the task type, load mode, or system deployment strategy, without the need for frequent major updates during the collaborative control process. During the same target task process, the system topology structure and efficiency are maintained as stable as possible, and it has good application and promotion prospects. Description of the Drawings

[0026] Figure 1 It is a schematic diagram of an agent node architecture in a multi-agent collaborative control network provided by an embodiment of the present invention;

[0027] Figure 2 It is a topological structure diagram of a collaborative control network provided by an embodiment of the present invention;

[0028] Figure 3 It is an architecture diagram of a control method of a collaborative control system provided by an embodiment of the present invention. Detailed Embodiments

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.

[0030] Accordingly, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0031] It should be noted that like reference numerals and letters denote like items in the following figures, and thus, once an item is defined in one figure, it will not be necessary to further define and explain it in subsequent figures.

[0032] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the inventive product is customarily placed during use. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention. In addition, the terms "first", "second", "third", etc. are only used for descriptive distinction and should not be construed as indicating or implying relative importance.

[0033] In addition, terms such as "horizontal" and "vertical" do not mean that the components are required to be absolutely horizontal or hanging, but may be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and does not mean that the structure must be completely horizontal, but may be slightly inclined.

[0034] In the description of the present invention, it should also be noted that unless otherwise clearly specified and limited, the terms "set", "installed", "connected", "connected" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] Although complex network systems have been verified to have many advantages in multiple practical scenarios, especially after combining multi-agent node devices, they can adjust the interaction mode in real time according to task requirements. For example, the synchronization control scheme under loose dynamic topology conditions can effectively cope with network structure changes and improve system flexibility. Many scholars have proposed a lot of intelligent algorithms to deal with traditional fields in this field. However, the relevant optimization schemes tend to be more theoretically guided; many problems have emerged in actual application scenarios after combining idealized schemes; one of the more fundamental problems among them is the connectivity problem of the collaborative system.

[0036] Under normal circumstances, the core of the motion coordination of the cooperative control network lies in constructing a connected multi-hop communication network to ensure the reliability and integrity of information exchange between agents. The multi-hop communication network allows agents to transmit information through multiple relay nodes, and connectivity requires that there is at least one communication path between any two agents in the network. Existing coordination algorithms usually assume that the network is always connected or jointly connected, but this assumption often goes wrong in actual application scenarios, and the reasons are diverse. For example, the perception and communication capabilities of agents have an upper limit, and the service life of components on physical entities is also limited, resulting in limited applicability of the algorithm. There are also differences in node radiation and carrying capacity in different scenarios. When these limitations and differences occur, it is extremely easy for the connectivity foundation of the cooperative control network to be impacted.

[0037] For the characteristics of agents in some specific scenarios (in complex cluster networks), if the connectivity of the cooperative control network is guaranteed regardless of the consequences, it will require a high computational cost, which is contrary to the original purpose of introducing system control; there are also solutions that propose decentralized algorithms to maintain neighbor connections by restricting the movement range of agents, but this relatively conservative strategy may also lead to an extreme, and the flexibility of the entire cooperative control system will be significantly reduced.

[0038] Referring to the appendix Figure 1 , the present invention proposes a multi-agent cooperative control system, including a plurality of agent nodes, and each agent node at least includes: a node intelligent processing unit (NIPU: Node Intelligent Processing Unit), a state perception and collection unit, a task execution and control unit, and a function monitoring and adjustment unit.

[0039] Among them, the node intelligent processing unit: It includes functions such as communication link data processing, status monitoring of adjacent processing units, collaborative preprocessing, fault problem analysis and evaluation, self-status learning, and system command control. It is composed of functional modules with functions such as communication, monitoring, and motion control, and supports data exchange within a limited communication radius; the NIPU can adjust the device and preset parameters according to the complexity of the processing task target in the actual scenario. Especially in scenarios with cost constraints, it is not necessary for all NIPU units to have high-intelligence processing functions. For example, local node intelligence can focus on NIPUs for execution control categories, while global node intelligence can focus on high-intelligence NIPUs for system control categories (usually with reinforcement learning modules and better hardware configuration).

[0040] The status perception and collection unit: It is mainly responsible for actively collecting various status information of this intelligent agent and its surrounding environment in real time, including position, speed, energy status, and sensor data. This unit provides necessary real-time data support for collaborative decision-making.

[0041] The task execution and control unit: According to the received task instructions and collaborative control strategies, it actively or passively executes specific actions. This unit includes task decomposition, motion instruction generation, and execution feedback modules to ensure action coordination and timely adjustment during the task process.

[0042] The function monitoring and regulation unit: It is used to monitor or regulate the energy consumption of the current intelligent agent, and adjust the radiation distribution of the intelligent agent in real time according to the working state to ensure the long-term stable operation of the system. For example, through dynamic power monitoring and regulation, load balancing and energy replenishment regulation between nodes are realized.

[0043] Each node area in the multi-agent collaborative control system composed of the above nodes corresponds to an intelligent agent. The intelligent agent can execute user instructions or tasks assigned by the upper layer, while monitoring its own operating state, collecting environmental feedback, and publishing heartbeats and status reports on the communication layer. Multiple intelligent agent nodes in the collaborative control system coordinate and cooperate with each other to jointly complete a certain complex task or goal.

[0044] Currently, the vast majority of connectivity maintenance control algorithms need to maintain all existing topological connections in the network. On the one hand, too many nodes will generate a lot of unnecessary redundant topological connections, limiting the movement range of the intelligent agent and reducing the flexibility of the collaborative control network; on the other hand, if the nodes are too sparse, although the computational overhead and communication cost are reduced, the connection stability is insufficient, and it is easy to be attacked and paralyzed. After improving the traditional collaborative control system, the overall network topology of the present invention is more elastic, more effective in dealing with problems such as complex task tracking and obstacle avoidance, and will not cause excessive control burden.

[0045] Taking obstacle avoidance in a swarm network as an example, the present invention first categorizes all agents in the collaborative system into nodes based on scenario requirements. These agents are then assigned to master control nodes, collaborative control nodes, and execution control nodes. Specifically, the master control node is selected based on the functional characteristics of the agents within the global system. Collaborative nodes are then selected based on an optimization algorithm, and execution nodes are assigned to form a collaborative control system network.

[0046] Global Control Node: The collaborative control network includes a global control node, which is mainly responsible for system initialization, task decomposition, resource allocation and global optimization strategy generation; Cooperative Control Node: It is used to receive task information from the global control node layer, coordinate tasks and path planning of intelligent agents in the local area, and has relay communication and local collaborative decision-making capabilities; Execution Control Node network: It is responsible for the execution of specific tasks, including mobile control, information collection and other operations, and cooperates with the global control node and cooperative control node to complete task indicators.

[0047] Reference Figure 2 In this example of a collaborative control network architecture in an embodiment of the present invention, agent A is the master control node, agents B, C, and D are collaborative control nodes, and the remaining agents are executive control nodes. The connections between the control nodes form a collaborative control network. To ensure the overall connectivity of the system, the connections between agents on the collaborative network are necessary and must be effectively maintained and controlled. The master control node must control and manage all subordinate collaborative control nodes, and the collaborative control nodes must control and manage all subordinate executive control nodes. Each dominant agent (master control node or collaborative control node) and the other agents under its control, along with the communication links between them, form a cluster.

[0048] In the solution of the present invention, the first step in building a collaborative control network is to screen out the master control intelligent agent nodes. The key role of the master control intelligent node is overall control and resource allocation. Although the collaborative control nodes and the execution control nodes can achieve a certain degree of multi-agent collaborative control, in order to achieve global situational awareness, cross-domain task scheduling, policy consistency control, and the fault tolerance and scalability of the system, it is necessary to design a master control node layer, which helps to improve the stability, flexibility and engineering feasibility of the entire collaborative control network. In particular, in the collaborative control system obstacle avoidance scenario, the problem of edge node interruption and failure that cannot be overcome well by general algorithms can be effectively solved.

[0049] In the solution of the present invention, the master control node is preferably screened by the functional characteristics of the intelligent node, more specifically, the radiation power adjustment performance, energy consumption, NIPU unit integrity (or intelligence) and node location of the intelligent node are considered.

[0050] The screening priority function of the master control node agent n is set to Z(n);

[0051] Z(n)=k1Z F (n)+k2Z P (n)+k3Z N (n)+k4Z L (n);

[0052] Among them, Z F (n), Z P (n), Z N (n), Z L (n) corresponds to the characterization functions of the intelligent agent's radiation power adjustment performance, the intelligent agent's energy consumption and power stability characteristics, the intelligent agent's node NIPU unit performance integrity, and the intelligent agent's node location. k1, k2, k3, and k4 are weight coefficients, which can be static settings, online learning, or scene adaptive settings.

[0053] The selection of master control nodes in the present invention takes these four key characteristics into consideration based on their actual scenario requirements within the collaborative network. This allows them to effectively compensate for the lags and deficiencies of the intelligent node collaborative algorithm. A higher-performance NIPU unit indicates a node with stronger learning capabilities and faster response speeds. By granting it higher authority, the master control node can achieve better overall network control performance, enabling timely assessment and evaluation of key aspects such as node addition and removal, fault handling, and complex obstacle avoidance path planning.

[0054] The radiation power adjustment feature is based on the stability of node network connections. In practice, perfect algorithms are rarely supported. Even the most perfect obstacle avoidance path design can still encounter unexpected situations (such as cluster network failures or attacks). These situations can easily cause intelligent nodes to fall out of the communication radius, affecting the achievement of collaborative control objectives. The master control node can make timely radiation adjustments to address these unexpected situations. If a node loses connection due to an obstacle during obstacle avoidance, the master control node can expand the radiation power radius to enhance the master control node's communication radius and achieve re-control of the lost node. If communication fails due to a local node attack, the master control node can establish a secondary connection using a backup radiation path, such as an alternative communication frequency band or type, to quickly restore the failed node's functionality.

[0055] Energy consumption stability, as a basic aspect of hardware operation, will also be taken into consideration. The consideration of node positions is mainly based on the self - safety lifespan and communication stability of the master control node, Z P (n) should preferably be an intelligent agent node located near the relatively central position in the physical space within the cooperative control network.

[0056] Regarding the selection of cooperative control nodes, although theoretically the method of using the Minimum Connected Dominating Set (MCDS) can more precisely identify the cooperative control nodes in the cooperative network, this method has great difficulties in actual engineering deployment. On the one hand, it often brings high computational complexity in large - scale networks. On the other hand, due to the need for global information or a large amount of message interaction, it is prone to problems such as control time delay and reduced real - time performance.

[0057] Based on this, the present invention proposes an improved solution that takes into account both practicality and computational efficiency, that is, by constructing an approximate connected dominating set as the cooperative control network. On the basis of introducing a small number of redundant nodes, this method uses a sub - optimal solution to replace the optimal solution, and significantly reduces the computational resources and time overhead required to solve the problem while trying to maintain the network control performance.

[0058] Specifically, the present invention uses a clustering algorithm based on multi - criterion priority ranking to achieve the selection and construction of cooperative control nodes. This algorithm constructs a priority weight system for cooperative task allocation by comprehensively evaluating multiple indicators such as the spatial position, node degree, communication radiation ability, and unique identity of each intelligent agent. On this basis, nodes with higher comprehensive weights are preferentially selected as cooperative control nodes, thereby completing the division of cooperative control nodes and execution control nodes in the cooperative control network.

[0059] Specifically, let Z(m) represent the priority of the cooperative control node of intelligent agent m. In order to achieve the unified ranking of multi - dimensional priority indicators, the priority function proposed by the present invention does not adopt the traditional weighted summation form, but uses a multi - index bit splicing mechanism, which is expressed as follows:

[0060] Z(m)=Z sp (m)⊕Z nd (m)⊕Z fq (m)⊕Z vp (m)⊕Z id (m);

[0061] Among them, ⊕ represents the bit - concatenation operation logic, which means that after converting each quantization index into a fixed - length binary code, they are concatenated in high - order priority according to the set order to construct a new fixed - length bit string. This bit string is finally used for node priority comparison and sorting.

[0062] Z sp (m) represents the proximity degree of agent m to the target obstacle. In practical applications, each agent uses its equipped sensing devices (such as sonar, lidar, etc.) to detect whether there are obstacles within its sensing range, thereby reducing the prior information that the agent obtains about the environment. This indicator occupies a high bit position in the screening priority of the cooperative control node, corresponding to a higher priority for this indicator; Z nd (m) represents the number of neighbors of agent m, characterizing the influence of the current node on other neighbors; Z fq (m) characterizes the communication quality between agent m and its neighbors and the quality of the radiation characteristics that the agent can provide, indicating that the current node has a better ability to handle possible communication problems and a longer system battery life, which can better ensure the stability of the system connection and promote the completion of tasks; Z id (m) is a priority indicator related to agent identity recognition to ensure the mutual exclusivity and uniqueness of agent priorities in the network.

[0063] According to the above priority definition method, the main cooperative control nodes in the cooperative network are screened out. Agents with higher priorities can dominate neighbor agents with lower priorities. According to this scheme, if an agent first detects the existence of an obstacle, its task priority may be higher than that of all its neighbor agents that have not detected the obstacle. Therefore, it is more likely to be selected as a cooperative control node, which is beneficial because usually the node that first discovers the obstacle is a key node that is closer to the obstacle (similarly, it can correspond to the faster progress of the agent in task processing, generally indicating that the agent has a more complete and high-performance NIPU unit), so as to initiate a rapid processing of this event; the number of neighbor nodes usually determines the sparsity of the network around the node. If a node has more neighbors, the permissions it needs to mobilize are relatively larger, and operations such as obstacle avoidance need to be carried out first; the radiation performance is still considered in the consideration of cooperative nodes. Although it does not reach the same level as in the master control node, the cooperative control node is still crucial for the entire cooperative control network because it usually has several execution networks under its jurisdiction. Therefore, if the radiation performance of the cooperative control node itself (such as performance affecting communication quality such as power, radiation coefficient, etc.) is average, it will inevitably increase the bandwidth consumption and computational burden of the master control node. Therefore, when screening cooperative control nodes, agents with better radiation performance should still be preferred. Finally, after screening out the master control node and the cooperative control nodes, the remaining system nodes are the execution control nodes, and the node allocation of the entire multi-agent system network is completed; the construction of the system network topology can be completed by combining the connection relationships between the nodes.

[0064] In practical applications, the ideal and desired redundant connections of the cooperative control network should be minimized as much as possible to reduce unnecessary communication constraints and motion constraints. The approximate minimum spanning tree method can be selected to extract the streamlined backbone subnet of the system and trim the redundant cooperative network connections. If three cluster heads form a triangle, the edge with the lowest priority is removed; if a gateway node is connected to multiple cluster heads simultaneously, the connection with the cluster head with the lowest priority connected to the gateway node is disconnected. This approach does not require global information and only needs local priority comparison to effectively reduce the number of connections.

[0065] After the network topology of the cooperative control system is completed, at the algorithm level, in the solution of the present invention, different potential functions are designed for composite processing to improve the success rates of connectivity maintenance, target task completion, obstacle avoidance, and collision avoidance in the multi-agent system.

[0066] In a preferred embodiment of the present invention, all agents in the system are directly or indirectly connected, and the connection relationship in the cooperative control nodes is reflected as the motion constraints between the agents of each cooperative control node. To maintain this constraint relationship, the agents of the cooperative control nodes should always be within the communication range of their associated neighbors (cooperative control nodes and the master control node) during the motion process. The motion of the executive control node agent is mainly controlled by the cluster head node agent it is associated with and moves following the cluster head agent. The master control node preselects to control the cooperative control node unit, but retains the authority to control all nodes. Especially when the edge node is interrupted due to obstacle avoidance or being attacked, the master control node needs to extend the communication radius / method to obtain control of the interrupted node and communication recovery.

[0067] To achieve the above coordinated control, the present invention constructs a composite artificial potential function, which at least includes the following four types of sub-potential functions: cooperative connectivity maintenance potential function U1, executive control attraction potential function U2, target attraction potential function U3, and obstacle avoidance repulsion potential function U4.

[0068] The composite potential function is defined as: U = μ1U1 + μ2U2 + μ3U3 + μ4U4, where μ is a preset scalar weight coefficient.

[0069] More specifically, the collaborative connection maintenance potential function U1 is used for the connection maintenance of collaborative control nodes in the system, to prevent the agents of the collaborative control nodes from losing communication connections with the neighbors of the collaborative control nodes or the master control node during movement; the execution control attraction potential function U2 is used to construct the connection maintenance between the execution control node and the leader node under its jurisdiction, and its connectivity is maintained by following the leader associated with it in the collaborative network. In general, the leader node is selected as the collaborative control node closest to the current execution control node as the leader node; the target attraction potential function U3 is an attraction potential function applied to the current agent based on the target position of each agent or the target position provided by the master control node; the obstacle avoidance repulsion potential function U4 is a repulsion potential function constructed to avoid collisions between interconnected agents or between agents and obstacles.

[0070] According to the content of the composite potential function, the collaborative control law of the node is given as:

[0071] = (µ1U1 + µ2U2 + µ3U3 + µ4U4)

[0072] Through the above, the multi-agent control system of the present invention divides the system network nodes specifically and sets the node control functions of the agents in the system specifically. Referring to Figure 3 , the multi-agent collaborative control method of the present invention is more specifically as follows:

[0073] The first step is the system initialization and startup. All agent nodes complete initialization according to the preset parameters carried by their own NIPU units, including communication parameter setting, attitude sensor calibration, communication neighborhood radius setting, etc.;

[0074] The second step is the construction of the collaborative control network. All agent nodes in the system are divided according to the set requirements; among them, the master control node is selected according to the radiation performance, energy consumption, integrity of the NIPU unit, and position weight of the agent node; the collaborative control node is selected according to the multi-index bit splicing mechanism; the rest are execution control nodes; the streamlined backbone subnet of the system is extracted according to the minimum spanning tree method, and the redundant collaborative network connections are trimmed to generate the collaborative control network;

[0075] The third step is to construct the control law of the node agent based on the composite potential function. The composite potential function at least includes the collaborative connection maintenance potential function U1, the execution control attraction potential function U2, the target attraction potential function U3, and the obstacle avoidance repulsion potential function U4;

[0076] Step 4: Dynamic monitoring and response of the collaborative control system. When a node becomes disconnected or goes out of the communication range after being attacked during the obstacle avoidance process of the collaborative control system, the master control node automatically triggers the radiation performance adjustment function, quickly recovers through hardware adjustment, and updates the collaborative control network structure in a timely manner.

[0077] Step 5: Periodic update of the collaborative control system. The above steps are cyclically executed within a preset control period to form an adaptive closed-loop control.

[0078] The above collaborative control method does not require complex collaborative control algorithms. It not only considers the connectivity of network nodes in complex situations but also takes into account the reduction of the computational consumption burden brought by using complex algorithms. It completes overall collaboration using a relatively convenient method. The priority function structure during the control process can be configured or fine-tuned according to the task type, load pattern, or system deployment strategy, without the need for frequent major updates during the collaborative control process. The system topology structure is kept stable as much as possible during the same target task process. It realizes the collaborative task execution, stable connectivity maintenance, and dynamic adaptive scheduling capabilities of multi-agent systems in complex environments.

[0079] To implement the above embodiments, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the program to implement the method for handling the failure of the main node of the train network as shown in the foregoing embodiments.

[0080] To implement the above embodiments, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. The program is characterized in that it is executed by a processor to implement the method for handling the failure of complex network nodes as shown in the foregoing embodiments. [[ID=1,4]]

[0081] To implement the above embodiments, the present invention also provides a computer program product. When the instructions in the computer program product are executed by a processor, the method for handling the failure of complex network nodes as shown in the foregoing embodiments is executed.

[0082] In addition, in order to implement the above-mentioned invention solutions, those skilled in the art can improve and supplement some detailed contents not specifically mentioned in the above solutions. For example, through the combination of artificial intelligence technologies such as deep learning and reinforcement learning, the collaborative decision-making and dynamic adaptation among multiple agents in the collaborative control network can be realized. The collaborative system can intelligently optimize the behavior strategies of each agent according to environmental changes and task requirements, improving the efficiency and stability of the overall system, etc. Another example is to improve appropriate algorithm functions to enhance the adaptive ability of the collaborative system, or to apply the basic concept of the present invention to the multi-agent collaborative control applications in fields such as unmanned systems, intelligent transportation, and intelligent manufacturing. These all belong to the reasonable extension under the basic concept of the present invention. If these related deformations do not depart from the main technical concept of this case, they should not be regarded as a breakthrough innovation to the invention creation of this case.

[0083] In summary, although the content of the present invention has been introduced in detail through the above preferred embodiments, it should be recognized that the above description should not be considered as a limitation of the present invention. After those skilled in the art have read the above content, various modifications and substitutions to the present invention will be obvious. Therefore, the protection scope of the present invention should be defined by the appended claims.

Claims

1. A multi-agent collaborative control system, including multiple agent nodes, characterized in that: Each of the agent nodes at least has a node intelligent processing unit, a state perception and collection unit, a task execution and control unit, and a function monitoring and adjustment unit; Each agent node of the collaborative control system is classified into node categories, specifically divided into: a master control node, a collaborative control node, and an execution control node. All nodes and the connection relationships of the nodes form a collaborative control network; The collaborative control system determines the collaborative control law of the nodes based on a composite potential function.

2. The multi-agent collaborative control system according to claim 1, characterized in that The function monitoring and adjustment unit is used to monitor the energy consumption of the current agent or adjust the radiation characteristics of the agent, and adjust the radiation distribution of the agent according to the real-time system working requirements.

3. The multi-agent collaborative control system according to claim 2, characterized in that, The selection of the master control node is determined based on the radiation power adjustment performance, energy consumption, integrity of the NIPU unit, and node position characteristics of the current agent node.

4. The multi-agent collaborative control system according to claim 3, characterized in that, The collaborative control node is selected and constructed based on a clustering algorithm with multi-criterion priority sorting for collaborative control nodes.

5. The multi-agent collaborative control system according to claim 4, characterized in that, This algorithm determines the priority function by comprehensively evaluating the spatial position, node degree, communication radiation ability, and unique identity index of each agent, as shown below: Z(m)=Z sp (m)⊕Z nd (m)⊕Z fq (m)⊕Z vp (m)⊕Z id (m); Among them, ⊕ represents the bit concatenation operation logic. After converting each quantization index into a fixed-length binary code, they are concatenated in high-order priority according to the set order to construct a new fixed-length bit string, which is finally used for node priority comparison and sorting.

6. The multi-agent collaborative control system according to claim 5, wherein After removing the master control node and the collaborative control node from the collaborative control system, the remaining ones are the execution control nodes.

7. The multi-agent collaborative control system according to claim 6, wherein The collaborative control system determines the collaborative control law of the agent nodes based on a composite artificial potential function.

8. The multi-agent collaborative control system according to claim 6, wherein If a lost node appears during the obstacle avoidance event of the collaborative control system, the master control node realizes the connection recovery of the lost node by adjusting the radiation characteristics of the current agent.

9. A control method for the multi-agent collaborative control system according to any one of claims 1-8, characterized in that It includes the following steps: The first step, the system is initialized and started. All agent nodes complete the initialization according to the preset parameters carried by their own NIPU units, including communication parameter setting, attitude sensor calibration, and communication neighborhood radius setting; The second step, the collaborative control network is constructed. All agent nodes in the system are divided according to the set requirements; The third step, based on the composite potential function, the control law of the node agent is constructed. The composite potential function at least includes a collaborative connection maintenance potential function U1, an execution control attraction potential function U2, a target attraction potential function U3, and an obstacle avoidance repulsion potential function U4; The fourth step, the dynamic monitoring and response of the collaborative control system. When a node is lost or detached from the communication range after being attacked during the obstacle avoidance process of the collaborative control system, the master control node automatically triggers the radiation performance adjustment function, quickly recovers through hardware adjustment, and updates the collaborative control network structure in a timely manner; The fifth step, the periodic update of the collaborative control system. The above steps are cyclically executed within a preset control period to form an adaptive closed-loop control.

10. The control method of the multi-agent collaborative control system according to claim 9, wherein For the construction of the collaborative control network, the master control nodes are selected according to the radiation performance, energy consumption, integrity of NIPU units, and position weights of the agent nodes; the collaborative control nodes are selected according to the multi-index bit splicing mechanism; the rest are execution control nodes; the streamlined backbone subnet of the system is extracted by the minimum spanning tree method, and the redundant collaborative network connections are trimmed to generate the collaborative control network.

Images