Decentralization swarm intelligence cooperative control method and system of robot system

Abstract

The invention provides a decentralized swarm intelligence cooperative control method and system of a robot system, belongs to the technical field of multi-robot cooperative control, and can at least partially solve the problems that an existing robot system is high in communication overhead; role distribution depends on central planning or assumes that global information is available; the formation model lacks a weighting mechanism for role behavior differences; in order to solve the problem of lack of a local robust strategy for failure recovery in the prior art, each robot constructs a neighbor table through local communication, calculates local fitness of different roles based on own state and environment information, and maintains a virtual pheromone matrix. And according to the pheromone matrix and the local fitness, roles are autonomously selected according to a probability transfer rule, and an expected motion control quantity is generated in combination with aggregation, separation and alignment rules so as to realize group formation control. Meanwhile, the pheromones are updated according to the task execution effect, role redistribution is triggered when the key role fails, and self-adaptive cooperative control over the robot group is achieved.

Description

Technical Field

[0001] This invention belongs to the field of multi-robot collaborative control technology, specifically relating to a decentralized swarm intelligent collaborative control method and system for robot systems. Background Technology

[0002] With the rapid growth of robotic systems applications in industrial collaboration, inspection and monitoring, warehousing and logistics, and space exploration, traditional centralized control frameworks are gradually revealing problems such as communication bottlenecks, single points of failure, scheduling delays, and poor scalability. When the environment is dynamically changing and partially observable, the central server struggles to integrate all robot information in a timely manner, leading to slow system response or even system failure.

[0003] Swarm intelligence methods offer a novel technological approach to solving the aforementioned problems by inspiring the distributed decision-making and collaborative behaviors exhibited by social organisms in nature (such as ants and flocks of birds). By enabling each robot to rely solely on local perception, neighbor communication, and simple rules, the self-organizing emergence of overall behavior can be achieved, significantly reducing reliance on a central server.

[0004] However, existing robot systems still have the following limitations: 1. High communication overhead, unsuitable for congested or restricted networks; 2. Role assignment often relies on centralized planning or the assumption that global information is available; 3. The formation model lacks a weighting mechanism for differences in role behavior; 4. Lacks local robust strategies for failure recovery. To address these issues, we propose a decentralized swarm intelligence cooperative control method and system for robot systems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and to provide a decentralized swarm intelligent collaborative control method and system for robot systems.

[0006] This invention provides a decentralized swarm intelligence collaborative control method for a robot system, comprising the following steps: S1: Each of the robots sends its own compressed status message at a preset period and receives compressed status messages from other robots. S2: Each robot constructs and updates its local neighbor table based on the received compressed status message. If it does not receive the compressed status message of a neighboring robot within a preset communication timeout period, it performs communication degradation processing on that robot. S3: Each robot calculates its local fitness for each role in the preset role set based on its own perception information, and maintains a virtual pheromone matrix corresponding to the preset role set. S4: Each robot autonomously selects its current role according to the probability transition rule based on the virtual pheromone matrix and the local fitness. S5: Each robot determines the cluster formation control parameters according to its current role, generates the desired motion control quantity, outputs the desired motion control quantity to the underlying motion controller of the robot system, and performs the task under the control of the underlying motion controller. S6: Each robot updates and strengthens the virtual pheromone matrix based on the task execution results.

[0007] Further, in step S1, the compressed status message includes the robot's identifier, timestamp, pose information, speed information, current role, and role pheromone vector. The data length of a single compressed status message does not exceed 150 bytes, and the preset period ranges from 10Hz to 50Hz.

[0008] Specifically, in step S2, the communication timeout period ranges from 3 seconds to 10 seconds, and the communication degradation process includes the following steps: Remove the corresponding neighbor robot from the local neighbor table and reduce the influence weight of the neighbor robot's information on the current robot's role selection and formation control.

[0009] Specifically, in step S3, the local fitness is calculated based at least on the distance from the robot to the target, the robot's remaining energy, and the robot's local risk, wherein the local risk includes at least one or more of collision risk, obstacle density, or communication packet loss rate; the fitness function corresponding to the local fitness is... ,in Distance to the target For energy, For localized risks, and Assign weights to them.

[0010] Preferably, in step S4, based on the virtual pheromone matrix and the local fitness, a probability transition rule based on pheromone values ​​and the local fitness is used to calculate the selection probability of each role, and random sampling is performed according to the selection probability to determine the current role; the probability selection function of the probability transition rule is... ,in, For the role pheromones, For role suitability, and Control the weight of both influences.

[0011] Specifically, in step S6, the volatilization and enhancement renewal includes the following steps: The pheromones of each character are attenuated according to a preset volatility coefficient, and the corresponding character's pheromones are enhanced based on the task effect after the robot performs the selected character's task.

[0012] Furthermore, when a communication interruption or failure is detected in the robot playing a key role, steps S3 to S6 are re-executed to complete the role reassignment and cause the robot swarm to converge again.

[0013] Another aspect of the present invention provides a decentralized swarm intelligent collaborative control system for a robot system, the system being adapted to implement the above-described decentralized swarm intelligent collaborative control method for a robot system, comprising: [the following components are provided on each of the robots]. A sensor module is used to acquire the robot's own motion state and environmental information; A state estimation module, electrically connected to the sensor module, is used to estimate the robot's own state based on the data output by the sensor module. A communication module, electrically connected to the state estimation module, is used to send and receive the compressed state message via UDP broadcast or multicast, and to maintain the local neighbor table; A swarm intelligence layer, electrically connected to the communication module, is used to perform distributed role selection based on the virtual pheromone matrix and the local fitness, and to generate the swarm formation control parameters; and A motion control module, electrically connected to the swarm intelligence layer, is used to convert the desired motion control quantity output by the swarm intelligence layer into control commands for the robot actuator. The swarm intelligence layer works in conjunction with the communication module to enable group collaboration among multiple robots without a central control node.

[0014] Furthermore, the swarm intelligence decision-making module includes a role allocation unit, an adaptive behavior layer, and a formation control unit; The role allocation unit performs role selection based on ant colony optimization principles, and the formation control unit generates control variables based on aggregation rules, separation rules, and alignment rules, combined with the role bias term; the function of the aggregation rule is... ,in, , For the robot's position, The aggregation weights are used; the function of the separation rule is... In this case, the square of the distance in the denominator makes the repulsion stronger for closer distances. Adjust its intensity; the alignment rule function is ,in, , For speed, To align the weights; the function of the control quantity is: ,in, This is a character bias.

[0015] Specifically, the swarm intelligence layer also includes a pheromone update unit and a fault detection and recovery unit, which are used to trigger role reassignment when a key role robot is detected to be offline, and to enable the robot group to re-form a cooperative state through the swarm formation control.

[0016] The beneficial effects of this invention are as follows: Compared with existing centralized or weakly distributed robot collaborative control schemes, the decentralized swarm intelligent collaborative control method and system for robot systems provided by this invention achieves autonomous collaboration of multiple robot groups without the need for a central control node through local communication, distributed role selection, and a collaborative decision-making mechanism based on virtual pheromones. This significantly improves the system's scalability, robustness, and task execution efficiency in complex environments. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the steps of a decentralized swarm intelligent collaborative control method for a robot system according to a specific embodiment of the present invention. Figure 2 A flowchart illustrating the role allocation process of a decentralized swarm intelligent collaborative control method for a robot system according to a specific embodiment of the present invention. Figure 3 This is a diagram illustrating the overall architecture of a decentralized swarm intelligent collaborative control system for a robot system according to a specific embodiment of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 , Figure 2 As shown in the figure, a decentralized swarm intelligent collaborative control method for a robot system provided by a specific embodiment of the present invention includes the following steps: S1: Each robot sends its own compressed status message at a preset period and receives compressed status messages from other robots. S2: Each robot builds and updates its local neighbor table based on the received compressed status messages. If a robot does not receive a compressed status message from a neighbor robot within the preset communication timeout period, it performs communication degradation processing on that robot. S3: Each robot calculates its local fitness for each role in the preset role set based on its own perception information, and maintains a virtual pheromone matrix corresponding to the preset role set; where the role set is... K types of roles can be customized according to the task; the fitness function is: ,in Distance to the target For energy, For localized risks, and Weighting is assigned based on proximity, energy level, and risk. Greater fitness is observed when distance is short, energy is high, and risk is low. S4: Each robot autonomously selects its current role based on the virtual pheromone matrix and its local fitness, according to a probability transition rule; the probability selection formula is as follows: ,in, For the role pheromones, For role suitability, and Control the weighting of these two factors. Higher pheromone levels and better fitness result in a greater probability of being selected. S5: Each robot determines the cluster formation control parameters according to its current role, generates the desired motion control quantity, outputs the desired motion control quantity to the underlying motion controller of the robot system, and executes the task under the control of the underlying motion controller. S6: Each robot updates the virtual pheromone matrix by evaporating and strengthening it based on the task execution results; the function for updating the pheromone is... ,in The volatility coefficient is... This contributes to new performance. The better the performance, the stronger the pheromones of the corresponding character.

[0020] Specifically, in step S2, the process of each robot constructing and updating its local neighbor table includes the following steps: Upon receiving a compressed status message from another robot, the robot identifier, timestamp, pose information, speed information, and current role information in the compressed status message are parsed; based on the robot identifier, a search is conducted in the local neighbor table to determine if a corresponding neighbor robot record already exists; if the neighbor robot record already exists, the latest status information in the compressed status message is used to update the corresponding record, and the communication timestamp of the neighbor robot is refreshed; if the neighbor robot record does not exist, a new neighbor robot record is created in the local neighbor table, and the status information and influence weight of the neighbor robot are initialized; the communication timestamps of each neighbor robot in the local neighbor table are periodically checked, and when the time difference between the current time and the communication timestamp exceeds a preset communication timeout, it is determined that the corresponding neighbor robot has experienced a communication anomaly; communication degradation processing is performed on the neighbor robot that has experienced a communication anomaly, removing it from the local neighbor table, or reducing the influence weight of its status information in subsequent role selection and formation control calculations; pheromones are locally estimated, allowing asynchronous and inconsistent behavior, and naturally converge through an evaporation mechanism.

[0021] Furthermore, by constructing and updating the local neighbor table, each robot can dynamically maintain the set of robots that it can perceive within its communication range without the need for global communication or a central node. This provides a reliable local state information basis for subsequent distributed role selection, group behavior coordination, and formation control calculations, and improves the robustness of the robot group in the event of communication instability or node failure.

[0022] Furthermore, in step S5, the process of generating the desired motion control quantity for each robot includes the following steps: based on the current role, select the aggregation weight, separation weight, alignment weight, and role bias term corresponding to the role from the preset role parameter set; based on the pose information of each neighboring robot in the local neighbor table, calculate the aggregation vector pointing to the geometric center of the neighboring robot according to the aggregation rule; based on the relative pose of each neighboring robot in the local neighbor table, calculate the separation vector to avoid close contact according to the separation rule; based on the velocity information of each neighboring robot in the local neighbor table, calculate the alignment vector to maintain motion consistency according to the alignment rule; weight the aggregation vector, separation vector, and alignment vector according to their corresponding weights and add the role bias term to obtain the desired motion direction vector; generate the desired linear velocity and desired angular velocity based on the desired motion direction vector to form the desired motion control quantity, and output it to the underlying motion controller; through the above steps, the abstract group cooperative behavior is transformed into directly executable motion control instructions, enabling the robot to maintain the stable structure of the group while performing differentiated behaviors according to its own role, realizing the organic combination of group cooperation and individual autonomous control.

[0023] Based on the above basic implementation method, in step S1, the compressed status message includes the robot's identifier, timestamp, pose information, speed information, current role, and role pheromone vector. The data length of a single compressed status message does not exceed 150 bytes, and the preset period range is 10Hz to 50Hz.

[0024] Specifically, in step S2, the communication timeout period is in the range of 3S to 10S, and the communication degradation process includes the following steps: removing the corresponding neighbor robot from the local neighbor table, and reducing the influence weight of the neighbor robot's related information on the current robot's role selection and formation control.

[0025] In one specific implementation, in step S3, the local fitness is calculated based at least on the distance from the robot to the target, the robot's remaining energy, and the robot's local risk, wherein the local risk includes at least one or more of collision risk, obstacle density, or communication packet loss rate.

[0026] In this embodiment, in step S3, each robot calculates its local fitness for each role in the preset role set. The process includes the following steps: acquiring the robot's own pose information, remaining energy information, and environmental perception information; calculating the distance index from the robot to the task target or reference position based on the pose information; calculating the local risk index of the area where the robot is located based on the environmental perception information, where the local risk index includes one or more of collision risk, obstacle density, or communication packet loss rate; normalizing the distance index, remaining energy index, and local risk index; and for each role in the preset role set, weighting and combining the normalized indices according to the performance requirements of the corresponding role to obtain the local fitness value corresponding to that role.

[0027] Furthermore, by introducing local fitness calculation, each robot can quantify its suitability for performing different roles based on its own state and environmental conditions. This enables differentiated and context-aware autonomous decision-making in subsequent role selection processes, avoiding energy waste or local risk accumulation caused by fixed role allocation and improving overall task execution efficiency.

[0028] In another specific embodiment, in step S4, based on the virtual pheromone matrix and local fitness, the selection probability of each role is calculated using a probability transition rule based on pheromone value and local fitness, and random sampling is performed according to the selection probability to determine the current role; in step S6, the evaporation and enhancement update includes the following steps: attenuating the pheromones of each role according to a preset evaporation coefficient, and enhancing the pheromones of the corresponding role according to the task effect after the robot performs the selected role; when it is detected that the robot undertaking the key role has communication interruption or failure, steps S3 to S6 are re-executed to complete the role reallocation and cause the robot group to converge again.

[0029] Specifically, the aforementioned probability transfer rule is used to introduce a probabilistic selection mechanism among multiple candidate roles, so that while the robot prioritizes roles with high pheromone values ​​and high local fitness, it still retains a certain degree of randomness, thereby avoiding the group from falling into a fixed role allocation or a local optimum.

[0030] In another specific embodiment, such as Figure 3 As shown, this invention provides a decentralized swarm intelligence collaborative control system for robot systems. This system is suitable for implementing the aforementioned decentralized swarm intelligence collaborative control method for robot systems. It includes: a sensor module installed on each robot for acquiring the robot's own motion state and environmental information; a state estimation module electrically connected to the sensor module for estimating the robot's own state based on the data output by the sensor module; a communication module electrically connected to the state estimation module for sending and receiving compressed state messages via UDP broadcast or multicast, and maintaining a local neighbor table; a swarm intelligence layer electrically connected to the communication module for performing distributed role selection based on a virtual pheromone matrix and local fitness, and generating swarm formation control parameters; and a motion control module electrically connected to the swarm intelligence layer for converting the desired motion control quantity output by the swarm intelligence layer into control commands for the robot's actuators. The swarm intelligence layer and the communication module work together to achieve group collaboration among multiple robots without a central control node. The communication module has a built-in neighbor robot communication link.

[0031] Furthermore, the swarm intelligence decision-making module includes a role allocation unit, an adaptive behavior layer, and a formation control unit. The role allocation unit performs role selection based on ant colony optimization principles, while the formation control unit generates control variables based on aggregation rules, separation rules, and alignment rules, combined with role bias terms. The weighted summation function of the three basic rules is as follows: Cohesion function: ,in, , For the robot's position, For aggregate weights; Separation function: In this case, the square of the distance in the denominator makes the repulsion stronger for closer distances. Adjust its intensity; Alignment function: ,in, , For speed, To align weights; Final control function: ,in, This is a character bias.

[0032] Furthermore, the swarm intelligence layer also includes a pheromone update unit and a fault detection and recovery unit, which are used to trigger role reassignment when a key role robot is detected to be offline, and to enable the robot group to re-establish a cooperative state through swarm formation control.

[0033] In one specific implementation, the system consists of the following modules: Sensors and local state estimators: including odometry, IMU, vision or depth sensors, etc., to estimate their own state vector si; Communication module: Periodically sends its own compressed state vector via UDP broadcast, receives neighbor data, and updates the local neighbor table; Swarm intelligence layer: includes ACO role assignment, flopping formation control, and adaptive behavior layer; The underlying motion controller maps the desired velocity and acceleration to joint or drive control quantities. In this embodiment, each robot periodically broadcasts the following compressed data packet: {RobotID, Timestamp, Pose(x,y,θ), Velocity(vx,vy), CurrentRole, PheromoneVector(K-dimensional)}. If no data is received from a neighbor for T consecutive seconds, then: the neighbor is removed from the local neighbor table; the dependency weight of the neighbor's pheromone is reduced; and the local exploration mode is automatically switched. In this mode, the size of the data packet does not exceed 150 bytes, and T ranges from 3S to 10S.

[0034] In this embodiment, the steps of the adaptive behavior layer and fault tolerance strategy are as follows: Local Enhancement Signal: Calculates real-time rewards based on mission progress, energy consumption, collision risk, etc., and updates character preferences; Fault detection: Communication timeout is marked as disconnection; Fault recovery: Automatically triggers a new round of ACO role election; flocking naturally reconverges; if the leader goes offline, the leader with the highest pheromone level will take over.

[0035] To aid in a better understanding of the present invention, a more comprehensive and specific embodiment is described, in which the present invention provides a decentralized swarm intelligence collaborative control method for a robot system, comprising the following steps: S1: Each robot sends its own compressed status message at a preset period and receives compressed status messages from other robots. S2: Each robot builds and updates its local neighbor table based on the received compressed status messages. If a robot does not receive a compressed status message from a neighbor robot within the preset communication timeout period, it performs communication degradation processing on that robot. S3: Each robot calculates its local fitness for each role in the preset role set based on its own perception information, and maintains a virtual pheromone matrix corresponding to the preset role set. S4: Each robot autonomously selects its current role based on the virtual pheromone matrix and local fitness, according to the probability transition rules; S5: Each robot determines the cluster formation control parameters according to its current role, generates the desired motion control quantity, outputs the desired motion control quantity to the underlying motion controller of the robot system, and executes the task under the control of the underlying motion controller. S6: Each robot updates and strengthens the virtual pheromone matrix based on the performance of the task.

[0036] In this embodiment, in step S1, the compressed status message includes the robot's identifier, timestamp, pose information, speed information, current role, and role pheromone vector. The data length of a single compressed status message does not exceed 150 bytes, and the preset period range is 10Hz to 50Hz. In step S2, the communication timeout time ranges from 3S to 10S, and the communication degradation processing includes the following steps: removing the corresponding neighbor robot from the local neighbor table and reducing the influence weight of the neighbor robot's related information on the current robot's role selection and formation control. In step S3, the local fitness is calculated based at least on the robot's distance from the target, the robot's remaining energy, and the robot's local risk, wherein the local risk includes at least one or more of collision risk, obstacle density, or communication packet loss rate. In step S4, based on the virtual pheromone matrix and local fitness, the selection probability of each role is calculated using a probability transition rule based on pheromone value and local fitness, and random sampling is performed according to the selection probability to determine the current role. In step S6, the evaporation and enhancement update includes the following steps: The pheromones of each role are attenuated according to a preset evaporation coefficient, and the corresponding role pheromones are enhanced based on the task effect of the robot after performing the selected role. When a communication interruption or failure is detected in a robot that plays a key role, steps S3 to S6 are re-executed to complete the role reallocation and cause the robot group to converge again.

[0037] Specifically, another aspect of the present invention provides a decentralized swarm intelligent collaborative control system for a robot system, which is suitable for implementing the above-described decentralized swarm intelligent collaborative control method for a robot system, including the following configuration on each robot: The system includes: a sensor module for acquiring the robot's own motion state and environmental information; a state estimation module electrically connected to the sensor module for estimating the robot's own state based on the data output by the sensor module; a communication module electrically connected to the state estimation module for sending and receiving compressed state messages via UDP broadcast or multicast and maintaining a local neighbor table; a swarm intelligence layer electrically connected to the communication module for performing distributed role selection based on a virtual pheromone matrix and the local fitness, and generating the swarm formation control parameters; and a motion control module electrically connected to the swarm intelligence layer for converting the desired motion control quantities output by the swarm intelligence layer into robot actions. The system provides control commands for the robot's movement. The swarm intelligence layer works in conjunction with the communication module to enable collaborative grouping of multiple robots without a central control node. The swarm intelligence decision-making module includes a role allocation unit, an adaptive behavior layer, and a formation control unit. The role allocation unit performs role selection based on ant colony optimization principles, while the formation control unit generates control variables based on aggregation, separation, and alignment rules, combined with role bias terms. The swarm intelligence layer also includes a pheromone update unit and a fault detection and recovery unit to trigger role reassignment when a critical role robot is detected to be offline, and to restore the collaborative state of the robot group through swarm formation control.

[0038] In summary, the embodiments disclosed herein have at least the following technical effects: In this invention, each robot completes information interaction by periodically broadcasting compressed status messages and maintaining a local neighbor table, without relying on a central control node or a global consistency synchronization mechanism. When some robots experience communication abnormalities or failures, the remaining robots can still independently complete role selection and formation control, avoiding the paralysis of the entire system due to a single point of failure. Since each robot only communicates and makes decisions with its local neighbors, the communication load increases locally rather than globally with the increase in the number of robots, making this method applicable to large-scale robot group collaboration scenarios and reducing the communication and computing pressure when the system expands. By employing compressed status messages and limiting message length, combined with communication timeout detection and communication degradation handling mechanisms, the robot system can operate stably under network conditions with communication packet loss, delay, or limited bandwidth, thereby improving the system's adaptability in actual deployment environments. The dynamic construction and updating mechanism of the local neighbor table can reflect the changes in the robot's state within the communication range in a timely manner, avoid using outdated or invalid neighbor information to participate in role selection and formation control calculations, thereby improving the accuracy of group behavior decisions; By introducing local fitness calculated based on factors such as target distance, remaining energy, and local risks, the robot can select a more suitable role according to its own capabilities and environmental conditions, avoiding energy waste or local risk concentration problems caused by fixed role allocation. Role selection is based on virtual pheromone matrix and probability transition rules, which allows the robot to prioritize the selection of better roles while retaining a certain degree of randomness, thereby enhancing the system's exploration capabilities and promoting dynamic adjustment of role allocation as the task progresses and the environment changes. Each robot determines different formation control parameters and role bias terms according to its selected role, so that while the group maintains overall structural stability, different robots can undertake different functions, which is conducive to completing complex collaborative tasks. By combining aggregation rules, separation rules, and alignment rules, the robot swarm can maintain reasonable spacing, directional consistency, and overall connectivity during movement, thereby reducing the risk of collisions and improving the stability of the swarm's movement. When a communication interruption or failure is detected in a robot playing a key role, the system can trigger role reallocation and reconverge the group structure, enabling the robot group to continue to perform tasks even under dynamic changes or local failures. By using pheromone evaporation and enhancement mechanisms, the decision-making process of the robot swarm can gradually reflect historical performance and current environmental conditions, enabling long-term adaptation to environmental changes and improving the overall quality of task completion.

[0039] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A decentralized swarm intelligence collaborative control method for a robot system, characterized in that, Includes the following steps: S1: Each of the robots sends its own compressed status message at a preset period and receives compressed status messages from other robots. S2: Each robot constructs and updates its local neighbor table based on the received compressed status message. If it does not receive the compressed status message of a neighboring robot within a preset communication timeout period, it performs communication degradation processing on that robot. S3: Each robot calculates its local fitness for each role in the preset role set based on its own perception information, and maintains a virtual pheromone matrix corresponding to the preset role set. S4: Each robot autonomously selects its current role according to the probability transition rule based on the virtual pheromone matrix and the local fitness. S5: Each robot determines the cluster formation control parameters according to its current role, generates the desired motion control quantity, outputs the desired motion control quantity to the underlying motion controller of the robot system, and performs the task under the control of the underlying motion controller. S6: Each robot updates and strengthens the virtual pheromone matrix based on the task execution results.

2. The decentralized swarm intelligent collaborative control method for a robot system according to claim 1, characterized in that, In step S1, the compressed status message includes the robot's identifier, timestamp, pose information, speed information, current role, and role pheromone vector. The data length of a single compressed status message does not exceed 150 bytes, and the preset period ranges from 10Hz to 50Hz.

3. The decentralized swarm intelligent collaborative control method for a robot system according to claim 1, characterized in that, In step S2, the communication timeout period ranges from 3 seconds to 10 seconds, and the communication degradation process includes the following steps: Remove the corresponding neighbor robot from the local neighbor table and reduce the influence weight of the neighbor robot's information on the current robot's role selection and formation control.

4. The decentralized swarm intelligent collaborative control method for a robot system according to claim 1, characterized in that, In step S3, the local fitness is calculated based at least on the distance from the robot to the target, the robot's remaining energy, and the robot's local risk, wherein the local risk includes at least one or more of collision risk, obstacle density, or communication packet loss rate; the fitness function corresponding to the local fitness is: ,in Distance to the target For energy, For localized risks, and Assign weights to them.

5. The decentralized swarm intelligent collaborative control method for a robot system according to claim 1, characterized in that, In step S4, based on the virtual pheromone matrix and the local fitness, a probability transition rule based on pheromone values ​​and local fitness is used to calculate the selection probability of each role, and random sampling is performed according to the selection probability to determine the current role; the probability selection function of the probability transition rule is... ,in, For the role pheromones, For role suitability, and Control the weight of both influences.

6. The decentralized swarm intelligent collaborative control method for a robot system according to claim 1, characterized in that, In step S6, the volatilization and enhancement renewal includes the following steps: The pheromones of each character are attenuated according to a preset volatility coefficient, and the corresponding character's pheromones are enhanced based on the task effect after the robot performs the selected character's task.

7. The decentralized swarm intelligent collaborative control method for a robot system according to any one of claims 1 to 6, characterized in that, When a communication interruption or failure is detected in the robot playing a key role, steps S3 to S6 are re-executed to complete the role reassignment and cause the robot group to converge again.

8. A decentralized swarm intelligence collaborative control system for a robot system, characterized in that, The system is adapted to implement a decentralized swarm intelligent collaborative control method for a robot system according to any one of claims 3 to 7, comprising: [the following components are provided on each of the robots] A sensor module is used to acquire the robot's own motion state and environmental information; A state estimation module, electrically connected to the sensor module, is used to estimate the robot's own state based on the data output by the sensor module. A communication module, electrically connected to the state estimation module, is used to send and receive the compressed state message via UDP broadcast or multicast, and to maintain the local neighbor table; A swarm intelligence layer, electrically connected to the communication module, is used to perform distributed role selection based on the virtual pheromone matrix and the local fitness, and to generate the swarm formation control parameters. as well as A motion control module, electrically connected to the swarm intelligence layer, is used to convert the desired motion control quantity output by the swarm intelligence layer into control commands for the robot actuator. The swarm intelligence layer works in conjunction with the communication module to enable group collaboration among multiple robots without a central control node.

9. The decentralized swarm intelligent collaborative control system for the robot system according to claim 8, characterized in that, The swarm intelligence decision-making module includes a role allocation unit, an adaptive behavior layer, and a formation control unit. The role allocation unit performs role selection based on ant colony optimization principles, and the formation control unit generates control variables based on aggregation rules, separation rules, and alignment rules, combined with the role bias term; the function of the aggregation rule is... ,in, , For the robot's position, The aggregation weights are used; the function of the separation rule is... In this case, the square of the distance in the denominator makes the repulsion stronger for closer distances. Adjust its intensity; the alignment rule function is ,in, , For speed, To align the weights; the function of the control quantity is: ,in, This is a character bias.

10. The decentralized swarm intelligent collaborative control system for the robot system according to claim 9, characterized in that, The swarm intelligence layer also includes a pheromone update unit and a fault detection and recovery unit, which are used to trigger role reassignment when a key role robot is detected to be offline, and to enable the robot group to re-establish a cooperative state through the swarm formation control.

Images