Distributed fault-tolerant control method and device for multi-robot system

Abstract

The invention provides a distributed fault-tolerant control method and device for a multi-robot system, and relates to the technical field of robots, and the method comprises the steps: constructing a system dynamics model containing an actuator fault and a sensor fault, designing an intermediate observer for each follower robot, and carrying out the fault-tolerant control of the robot. The joint real-time estimation of the system state, the actuator fault, the sensor fault and the intermediate variable is realized, the estimation process does not need to know the boundary information of the fault amplitude and the change rate in advance, and does not need to meet the matching condition depended by a traditional observer, so that the adaptability and robustness of fault estimation are remarkably improved; on the basis, external state information of a leader and a neighbor robot is obtained from a communication network, a distributed cooperative fault-tolerant controller is constructed to generate an action control instruction, and the adverse effect of faults of an actuator and a sensor on the cooperative performance of the system is effectively compensated. And finally, all follower robots can stably and consistently track the motion state of the leader.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robots, in particular to a distributed fault-tolerant control method and device for a multi-robot system. BACKGROUND

[0002] Multi-robot systems are widely used in complex tasks such as cooperative rescue and cooperative combat, but in the actual operation process, they are inevitably affected by environmental interference and physical structure limitations, resulting in faults of actuators and sensors, which seriously affect the cooperative performance and stability of the system. Existing fault estimation methods such as sliding mode observer, adaptive observer, unknown input observer and high gain observer can estimate the faults of sensors and actuators to a certain extent, but there are several key defects: first, the upper bound of the fault function and its first derivative is usually required to be known, which is difficult to meet in actual application; second, it depends on the strict observer matching condition, that is, the fault gain matrix needs to satisfy a specific structural constraint, which limits the universality of the method; third, most existing schemes cannot estimate the system state and fault state simultaneously, and the design process is complex and tedious, which is not conducive to deployment on resource-limited multi-robot platforms. SUMMARY

[0003] The present application provides a distributed fault-tolerant control method and device for a multi-robot system to solve one or more technical problems existing in the prior art, at least to provide a beneficial choice or to create conditions, which can realize the simultaneous and accurate estimation of system state, actuator fault and sensor fault without relying on the upper limit of the amplitude and the change rate of the actuator and sensor faults, and without satisfying the observer matching condition, and effectively compensate for the influence of faults on the cooperative performance of multi-robots.

[0004] In one aspect, the present application provides a distributed fault-tolerant control method for a multi-robot system, the multi-robot system comprising one leader robot and a plurality of follower robots, each of the follower robots obtaining state information of the leader robot and neighbor robots through a communication network as external state information; The method comprises the following steps: Constructing a dynamic model of the multi-robot system, which includes actuator faults and sensor faults; For each follower robot, an intermediate observer is constructed by introducing an intermediate variable for relating the system state and the actuator fault, in combination with the output information of the dynamic model; Generating a joint estimation value of the system state, the actuator fault, the sensor fault and the intermediate variable in real time through the intermediate observer; According to the joint estimation value and the external state information, a distributed cooperative fault-tolerant controller is constructed in combination with the communication topology to generate an action control instruction; applying the motion control instruction to actuators of corresponding follower robots, so that all follower robots achieve consistent tracking of the motion state of the leader robot.

[0005] Further, the constructing the dynamic model of the multi-robot system comprises the following steps: describing the motion state of each robot as a system state containing position and velocity; modeling the actuator fault as an unknown abnormal bias acting on the robot control input channel, for representing actual fault situations such as actuator performance degradation, sticking or failure, etc. modeling the sensor fault as an unknown disturbance superimposed on the robot output signal, for reflecting actual measurement abnormalities such as measurement noise surge, sensor drift or data loss, etc. based on the dynamic correlation between the system state, control input, actuator fault and sensor fault, establishing a dynamic model reflecting the dynamic characteristics of robot motion.

[0006] Further, the constructing, for each follower robot, an intermediate observer by introducing an intermediate variable for correlating the system state with the actuator fault, in combination with the output information of the dynamic model, comprises the following steps: constructing an intermediate variable for each follower robot, which is determined by a linear combination relationship between the actuator fault and the system state; based on the intermediate variable and the output information of the dynamic model, establishing a local observer dynamic equation to form an intermediate observer structure suitable for a single follower robot; interconnecting the local intermediate observers of each follower robot through a communication network to form a distributed intermediate observer.

[0007] Further, the generating, by the intermediate observer, a joint estimation value of the system state, actuator fault, sensor fault and intermediate variable in real time comprises the following steps: generating, by the intermediate observer, the estimated output information of each follower robot according to its internal dynamic structure; calculating the deviation between the actual output information and the estimated output information of the follower robot, and inputting the deviation into the dynamic equation of the intermediate observer to update its internal state; based on the updated internal state of the intermediate observer, synchronously outputting the estimation values of the system state, actuator fault, sensor fault and intermediate variable to form the joint estimation value.

[0008] Further, the method further comprises: obtaining system state estimation and intermediate variable estimation of each follower robot; the intermediate variable estimation is used to equivalently represent actuator faults; receiving external state information from neighbor robots and leader robots; determining information connection relationship between each follower robot and its neighbors according to the communication topology; calculating state deviation of each follower robot relative to its neighbor robots and leader robot as cooperative error based on the system state estimation, the external state information and the communication topology; combining the cooperative error with the intermediate variable estimation, and performing amplitude limiting and smoothing processing on the combination result to generate the action control instruction.

[0009] Further, the distributed cooperative fault-tolerant controller determines the specific form of the action control instruction based on the preset cooperative control target and the robot platform type before generating the action control instruction: for wheeled mobile robots, the action control instruction includes linear velocity and angular velocity; for legged robots or mechanical arms, the action control instruction includes joint angle and joint torque; for unmanned aerial vehicles, the action control instruction includes pitch angle, roll angle, yaw angle and throttle control amount.

[0010] Further, the method further comprises: determining the number of follower robots and the communication topology type according to the actual task scenario; when the number of follower robots is less than a preset threshold and the communication environment is good, adopting undirected graph topology to ensure the symmetry and integrity of information interaction; when the number of follower robots exceeds the preset threshold or the communication bandwidth is limited, adopting directed graph topology to reduce network load; Each robot obtains its own state information through an external positioning system: in an outdoor environment, a global satellite positioning system is adopted; in an indoor environment, an ultra-wideband, infrared or visual motion capture system is adopted; at the same time, through a wireless communication module, each robot periodically exchanges position, velocity, orientation and remaining power information with neighbor robots.

[0011] Further, the method further comprises: each follower robot obtains its own state output information through an inertial measurement unit, a laser radar, a visual sensor or an encoder; In a normal working condition, the self-state output information is input to the intermediate observer as actual output information; when there is a sensor fault, the intermediate observer estimates the type, size and occurrence time of the sensor fault in real time based on the input-output relationship of the dynamic model and historical data trends, and fuses the estimated sensor fault with the partially reliable self-state output information to generate a compensated system state estimation value as part of the joint estimation value for use by the distributed collaborative fault-tolerant controller.

[0012] Further, the method further comprises: before applying the action control instruction to the actuator of the corresponding follower robot, judging whether the actuator has failed based on the current, voltage, output torque and response speed parameters of the actuator, in combination with the preset health state threshold and dynamic performance index; if no failure occurs, directly executing the action control instruction and feeding back the execution state to the intermediate observer to update the system output information; if an actuator failure is detected, estimating the type, degree and executable ability of the actuator failure in real time based on the dynamic model of the actuator and the actual output data by the intermediate observer, and using the estimated actuator failure to correct the intermediate variable estimation value, so as to realize compensation for the actuator failure when generating subsequent action control instructions.

[0013] In another aspect, the application provides a computer device deployed in each robot in a multi-robot system, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the distributed fault-tolerant control method of the multi-robot system when executing the computer program.

[0014] The application has the following beneficial effects: the application provides a distributed fault-tolerant control method for a multi-robot system, which constructs a system dynamic model containing actuator failure and sensor failure, and designs an intermediate observer based on intermediate variables for each follower robot, thereby realizing joint real-time estimation of system state, actuator failure, sensor failure and intermediate variables, without needing to know the boundary information of fault amplitude and its change rate in advance, and without needing to satisfy the matching condition relied on by traditional observers, thereby significantly improving the adaptability and robustness of fault estimation; on this basis, the external state information of the leader and neighbor robots obtained from the communication network and the communication topology structure of the system are combined to construct a distributed collaborative fault-tolerant controller to generate action control instructions, effectively compensating for the adverse effects of actuator and sensor failure on the collaborative performance of the system, and ultimately ensuring that all follower robots can stably and consistently track the motion state of the leader. The application also provides a corresponding device, which has similar beneficial effects as the method, and will not be described here.

[0015] Other features and advantages of the present application will be set forth in the description that follows, and in part will be apparent from the description, or can be learned by practice of the application. The purposes and other advantages of the present application will be realized and attained by the structure particularly pointed out in the description and claims hereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and are intended to serve as an exemplification of the application, together with the description, and are not intended to limit the application in its broader aspects.

[0017] Figure 1 is a flow chart of a distributed fault-tolerant control method of a multi-robot system provided by the present application; Figure 2 is a structural diagram of a distributed fault-tolerant control device of a multi-robot system provided by the present application. DETAILED DESCRIPTION

[0018] In order to make the purposes, technical solutions and advantages of the present application more clear, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and do not limit the present application.

[0019] The present application is further described below in combination with the drawings and specific embodiments. The described embodiments should not be considered as limiting the present application, and all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present application.

[0020] In the following description, "some embodiments" are described, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application, and are not intended to limit the present application.

[0022] In the research and application of multi-robot systems, cooperative control technology is the core support to achieve swarm intelligence and complete complex tasks. Such systems are widely used in cooperative rescue, cooperative combat, intelligent warehousing, environmental monitoring, and automated manufacturing scenarios. The key goal is to enable multiple robots to efficiently and stably complete designated tasks under conditions of limited information interaction and dynamic environmental changes. However, in actual operation, multi-robot systems are inevitably affected by factors such as internal component aging, external interference, or communication interruption, leading to faults in actuators and sensors. Actuator faults may manifest as a decrease in output torque, sticking, or even complete failure, while sensor faults may manifest as measurement drift, data loss, or a sudden increase in noise. If these faults are not identified and compensated for in a timely manner, they will seriously undermine the cooperative consistency of the system and even cause the overall task to fail or the system to collapse.

[0023] To address the above problems, existing technologies have developed various fault estimation and fault-tolerant control methods. Among them, the more typical ones include sliding mode observers, adaptive observers, unknown input observers, and high-gain observers. The basic idea of these methods is to detect faults by constructing system residual signals and further estimate the type, amplitude, and occurrence time of the faults to provide compensation for the controller. For example, sliding mode observers use discontinuous control laws to achieve robust estimation of faults; adaptive observers approximate fault signals by adjusting parameters online; unknown input observers attempt to decouple and estimate faults as unknown inputs of the system; and high-gain observers amplify observation errors to speed up fault response. These methods have achieved certain results in single-machine systems or idealized simulation environments.

[0024] However, when applying these existing techniques to practical multi-robot systems, several fundamental defects are exposed. First, most existing methods assume that the magnitudes of actuator faults and sensor faults and their first-order derivatives have known upper bounds in the theoretical design phase, i.e., the rate and strength range of fault changes need to be known in advance. This assumption is often difficult to establish in real-world scenarios because the occurrence of faults is sudden and uncertain, and their dynamic characteristics cannot be accurately predicted in advance. Second, existing observer designs generally rely on so-called observer matching conditions, which require the fault gain matrix to satisfy certain structural constraints, such as full column rank or having some algebraic relationship with the system output matrix. This condition is particularly difficult to satisfy in multi-robot systems because the dynamics models of different robots may differ, and the communication topology structure changes dynamically, resulting in the overall system not having a unified matching structure. Third, most existing schemes cannot simultaneously estimate the system state, actuator faults, and sensor faults in the same framework, often requiring the design of multiple independent observers, which not only increases the computational burden but also makes information fusion complex and coordination difficult. In addition, these methods usually have a cumbersome design process and rely on a large amount of prior knowledge for parameter tuning, making it difficult to efficiently deploy on resource-constrained embedded robot platforms, limiting their engineering practicality.

[0025] In summary, the current fault estimation and fault-tolerant control techniques for multi-robot systems have obvious deficiencies in theoretical assumptions, structural constraints, estimation capabilities, and engineering deployability. To address these issues, the present application proposes a distributed fault-tolerant control method and device based on an intermediate observer. By introducing an intermediate variable that relates the system state to the actuator fault in each follower robot, an intermediate observer is constructed that does not rely on prior knowledge of fault magnitudes and their rate of change, nor does it need to satisfy traditional observer matching conditions, thereby enabling real-time joint estimation of system state, actuator faults, sensor faults, and intermediate variables. Based on this, a distributed collaborative fault-tolerant controller is designed by combining the external state information of the leader and neighbor robots obtained from the communication network and the communication topology structure of the system, generating motion control instructions that can effectively compensate for fault effects, ultimately ensuring that all follower robots can stably and consistently track the motion state of the leader even in the presence of actuator and sensor faults, significantly improving the robustness, adaptability, and collaborative control performance of multi-robot systems in complex and uncertain environments.

[0026] First, the distributed fault-tolerant control method of the multi-robot system provided by the embodiments of the present application will be described in detail below with reference to the accompanying drawings. The multi-robot system includes a leader robot and multiple follower robots, and each follower robot obtains the state information of the leader robot and neighbor robots through a communication network as external state information, thereby providing necessary external state input for subsequent cooperative control. This information sharing method based on communication topology not only supports the implementation of distributed control, but also ensures that the system can still maintain cooperative consistency under partial connection or limited communication conditions, which is a basic prerequisite for realizing consistent tracking and fault-tolerant cooperative control.

[0027] Referring to Figure 1 The implementation process of the distributed fault-tolerant control method of the multi-robot system provided by the embodiments of the present application includes but is not limited to the following steps.

[0028] Step S110, constructing a dynamics model of the multi-robot system, including actuator faults and sensor faults.

[0029] In step S110, a mathematical description framework that can truly reflect the running characteristics of the system is established for the entire fault-tolerant control method. By explicitly introducing actuator faults and sensor faults in the dynamics model, the model not only covers the normal motion state of the robot, but also depicts the influence of the two key hardware faults on the dynamic behavior of the system. This modeling method provides a theoretical basis for subsequent fault estimation and compensation, so that the controller design can directly face the actual system containing fault factors, thereby improving the practicality and robustness of the method.

[0030] Step S120, for each follower robot, an intermediate observer is constructed by introducing an intermediate variable for associating the system state and the actuator fault, in combination with the output information of the dynamics model.

[0031] In step S120, an internal relationship between the system state and the actuator fault is established by constructing a special intermediate variable, and a local observer structure suitable for a single follower robot is designed based on this. The intermediate variable serves as a bridge connecting the state and the fault, so that the actuator fault, which is originally difficult to directly observe, can be indirectly reconstructed through the system output information. The intermediate observer constructed in combination with the output information of the dynamics model provides a feasible structural support for realizing joint estimation of faults and states.

[0032] Step S130, generating real-time joint estimation values of the system state, the actuator fault, the sensor fault and the intermediate variable through the intermediate observer.

[0033] The generation process of the joint estimation values does not depend on the prior knowledge of the upper limit of the actuator fault and the sensor fault and the change rate thereof, and does not require the multi-robot system to satisfy the observer matching condition.

[0034] In step S130, the internal dynamic mechanism of the intermediate observer is utilized to synchronously output real-time estimation results of the system state, actuator faults, sensor faults, and intermediate variables. These estimation values collectively constitute a joint estimation vector, which provides comprehensive and consistent information input for the subsequent controller. In particular, this process does not rely on prior knowledge of the upper limit of fault amplitude or its rate of change, nor does it require the system to satisfy the matching condition commonly seen in traditional observer design, significantly reducing the method's assumption requirements for fault characteristics and enhancing its applicability in actual complex environments.

[0035] Specifically, the generation process of the joint estimation values in this application does not rely on prior knowledge of the upper limit of the actuator fault and sensor fault amplitude and its rate of change, because the intermediate observer structure adopted by it embeds fault information into the reconfigurable auxiliary dynamics by introducing an intermediate variable that is a linear combination of the system state and the actuator fault, thereby avoiding explicit modeling or assuming the fault derivative or boundary. At the same time, this method uses the deviation feedback mechanism between the output information of the dynamic model and the internal state of the observer to adaptively drive the estimation error to converge, without prior knowledge of the specific form or rate of change of the fault. In addition, since the design of the intermediate variable decouples the strong algebraic constraints between the fault and the system output matrix, the entire observer structure no longer requires the fault gain matrix to satisfy the observer matching condition necessary for traditional methods, making this scheme applicable to a wider range of multi-robot system structures and significantly enhancing the generality and engineering applicability of fault estimation.

[0036] In step S140, a distributed cooperative fault-tolerant controller is constructed based on the joint estimation values and external state information, combined with the communication topology structure, to generate action control instructions to compensate for the adverse effects of actuator faults and sensor faults on the cooperative performance of the multi-robot system.

[0037] In step S140, the joint estimation values obtained in the previous step and the external state information obtained from the communication network are utilized to design a control strategy that can actively counteract the effects of faults, taking into account the system communication topology structure. The distributed cooperative fault-tolerant controller constructed by fusing local estimates and neighbor information generates targeted action control instructions, effectively weakening or eliminating the interference of actuator and sensor faults on the cooperative consistency of multi-robots, thereby ensuring that the overall cooperative performance of the system does not significantly degrade due to local faults.

[0038] In step S150, the action control instructions are applied to the actuators of the corresponding follower robots, enabling all follower robots to achieve consistent tracking of the motion state of the leader robot.

[0039] In step S150, the final execution link of the control loop is completed, i.e., the generated action control instruction is actually applied to the actuators of each follower robot to drive the follower robot to operate according to the compensated control law. Through this operation, even in the case of actuator or sensor failure, all follower robots can still continuously adjust their motion states, and ultimately achieve stable and consistent tracking of the leader robot trajectory or state, achieving the fundamental goal of multi-robot system cooperative control.

[0040] In some embodiments of the present application, in step S110, a dynamic model of the multi-robot system is constructed, including the following steps.

[0041] In step S210, the motion state of each robot is described as a system state including position and velocity.

[0042] In step S210, the basic state variable composition of each individual in the multi-robot system is determined, and by taking position and velocity as the core components of the system state, a mathematical foundation is established that can completely describe the motion behavior of the robot. This state description method not only conforms to the actual characteristics of the physical system, but also facilitates subsequent modeling, observation and control design, providing a clear and operable state space expression for the entire fault-tolerant control method.

[0043] In step S220, the actuator failure is modeled as an unknown abnormal deviation acting on the robot control input channel, which is used to represent actual failure situations such as actuator performance degradation, sticking or failure.

[0044] In step S220, various typical faults that may occur in the actuator are uniformly mathematically abstracted, and are regarded as unknown deviation signals superimposed on the control input. This modeling method can effectively cover actual situations such as reduced output capability or complete loss of output capability of the actuator due to aging, mechanical sticking or drive failure, etc., so that the dynamic model has the inclusiveness of actuator failure, and provides an accurate fault action path description for subsequent fault estimation and compensation.

[0045] In step S230, the sensor failure is modeled as an unknown disturbance superimposed on the robot output signal, which is used to reflect actual measurement abnormalities such as measurement noise surge, sensor drift or data loss.

[0046] In step S230, various abnormal situations that the sensor may encounter in actual operation are uniformly represented as unknown disturbances on the output channel. In this way, whether it is a surge in noise caused by environmental interference, a reading drift caused by device aging, or a data loss caused by communication interruption, it can be reasonably included in the model level, so as to ensure that the uncertainty of the system output information can be reflected within the dynamic framework, laying a modeling foundation for subsequent identification and compensation of sensor failure.

[0047] Step S240, based on the dynamic correlation between system state, control input, actuator fault and sensor fault, a dynamic model reflecting the dynamic characteristics of robot motion is established.

[0048] In step S240, the aforementioned elements are integrated to construct a unified dynamic model that can comprehensively describe the dynamic behavior of the robot under normal and fault conditions. This model not only contains the internal law of the evolution of system state over time, but also explicitly reflects the influence mechanism of control input, actuator fault and sensor fault on the input-output relationship of the system. By establishing this complete dynamic correlation structure, an accurate, consistent and physically meaningful system description basis is provided for subsequent observer design and fault-tolerant controller implementation.

[0049] In some embodiments of the present application, in step S120, for each follower robot, an intermediate observer is constructed by introducing an intermediate variable for associating the system state and the actuator fault, combined with the output information of the dynamic model, including the following steps.

[0050] Step S310, for each follower robot, an intermediate variable is constructed, which is determined by the linear combination relationship of the actuator fault and the system state.

[0051] In step S310, by defining a special intermediate variable, an explicit mathematical relationship between the actuator fault and the system state is established. The intermediate variable is not a physically measurable signal, but an auxiliary construction quantity, and its design purpose is to embed the actuator fault information that is difficult to separate directly into the structure related to the system state, so as to provide available algebraic or dynamic relationships for subsequent observer design, so that the fault information can be indirectly reconstructed through the system observable output.

[0052] Step S320, based on the intermediate variable and the output information of the dynamic model, the local observer dynamic equation is established to form an intermediate observer structure suitable for a single follower robot.

[0053] In step S320, the correlation between the intermediate variable and the actual output of the system is used to construct a dynamic observer model that only depends on local information. The local observer dynamic equation takes the system output as input, and through the evolution of the internal state, it approximates the system state and the intermediate variable in real time, thereby providing a basis for fault estimation. This structure is designed for a single follower robot and does not depend on global information, which embodies the localization and scalability of the method and is a key link to realize distributed fault estimation.

[0054] Step S330, interconnect the local intermediate observers of each follower robot through the communication network to form a distributed intermediate observer.

[0055] In step S330, the individual local intermediate observers are connected through the existing communication network of the multi-robot system to form a whole collaborative distributed observation architecture. This interconnection is not simply superimposed, but through the exchange of estimated information between neighbors, each observer can fuse neighborhood knowledge while maintaining local computing ability, improving estimation accuracy and robustness. The distributed intermediate observer thus constructed not only retains the efficiency of local processing, but also has the collaborative advantage of group cooperation, supporting the stable operation of the entire system under partial communication or local failure.

[0056] In some embodiments of the present application, in step S130, the joint estimated values of the system state, actuator faults, sensor faults and intermediate variables are generated in real time by the intermediate observer, including the following steps.

[0057] In step S410, the estimated output information of each follower robot is generated by the intermediate observer according to its internal dynamic structure.

[0058] In step S410, the local dynamic model constructed by the intermediate observer is used to deduce an estimated output signal corresponding to the real robot output based on its current internal state and known system structure. This estimated output reflects the best prediction of the system behavior under the current observer state and is the basis for subsequent error calculation and state correction, ensuring that the observer can continuously track the actual system trajectory.

[0059] In step S420, the deviation between the actual output information of the follower robot and the estimated output information is calculated, and the deviation is input into the dynamic equation of the intermediate observer to update its internal state.

[0060] In step S420, the actual output information of the robot By comparing the actual measurable output with the estimated output generated by the observer, the deviation signal reflecting the difference between the model and the real system is obtained, and the deviation is injected as a feedback term into the dynamic equation of the intermediate observer. This closed-loop correction mechanism drives the observer internal state to continuously approach the real system state, thereby realizing the adaptive adjustment and accurate approximation of the system dynamics and fault information.

[0061] In step S430, based on the updated internal state of the intermediate observer, the estimated values of the system state, actuator faults, sensor faults and intermediate variables are output synchronously to form the joint estimated values.

[0062] In step S430, all necessary information is extracted from the intermediate observers that have completed state updates, and real-time estimation results for system state, actuator faults, sensor faults, and intermediate variables are generated synchronously in one go. These estimates together constitute a complete joint estimation vector, providing unified, coordinated, and highly consistent information input for subsequent fault-tolerant controllers. This is a key step in achieving seamless integration of fault perception and collaborative control.

[0063] In some embodiments of this application, step S140 involves constructing a distributed collaborative fault-tolerant controller based on the joint estimate and external state information, combined with the communication topology, and generating action control commands, including the following steps.

[0064] Step S510: Obtain the system state estimate and intermediate variable estimate for each follower robot. The intermediate variable estimate is used to equivalently characterize actuator failure.

[0065] In step S510, two key pieces of information are extracted from the output of the intermediate observer: one is the estimate of the robot's current motion state, and the other is the estimate of the intermediate variables. Since the intermediate variables are constructed by a linear combination of the system state and the actuator fault, their estimated values ​​can effectively reflect the impact of the actuator fault, thereby achieving an equivalent characterization of the fault effect without directly measuring the fault, and providing the necessary basis for fault compensation for the subsequent controller.

[0066] Step S520: Receive external status information from the neighboring robot and the leader robot.

[0067] In step S520, each follower robot can acquire state data of other robots in its neighborhood and the leader through the communication network, including key information such as position and speed. This external state information is the basic input for realizing multi-robot cooperative control, ensuring that each individual not only relies on its own perception when making decisions, but also integrates into the group context, thereby supporting the achievement of the consistent tracking goal.

[0068] Step S530: Determine the information connection relationship between each follower robot and its neighbors based on the communication topology.

[0069] In step S530, the structural framework of information flow in the system is defined, i.e., which robots can exchange state information with each other. The communication topology determines the information basis of cooperative control and directly affects the calculation range of cooperative error and the distribution characteristics of control commands. By accurately identifying the connection relationship between each follower and its neighbors, it can be ensured that the controller is designed only based on actually accessible information, guaranteeing the applicability and stability of the method under different communication conditions.

[0070] Step S540: Based on the system state estimate, external state information, and communication topology, calculate the state deviation of each follower robot relative to its neighboring robots and the leader robot as the cooperative error.

[0071] In step S540, the locally estimated state is compared with the state information received from neighbors and the leader. Combined with the connection relationships defined by the communication topology, the deviation between the current robot's behavior and the group's expected behavior is quantified. This cooperative error reflects the system's current performance gap in consistency, serves as the direct basis for driving the fault-tolerant controller to generate correction instructions, and is also a key intermediate quantity for achieving distributed cooperative control.

[0072] Step S550: Combine the collaborative error with the estimated value of the intermediate variable, and perform amplitude limiting and smoothing on the combined result to generate motion control commands.

[0073] In step S550, the coordination error reflecting the group's coordination needs is fused with the estimated value of the intermediate variable characterizing the impact of actuator failure to form a control signal that comprehensively considers the coordination objective and fault compensation. Subsequently, by applying amplitude limiting and smoothing processing to this combined signal, the control commands are prevented from exceeding the physical capabilities of the actuator or causing violent actions, while also improving the continuity and safety of the commands, ultimately outputting motion control commands suitable for actual execution.

[0074] In some embodiments of this application, considering the collaborative control process of a multi-robot system, each robot may be affected by actuator or sensor malfunctions. These internal malfunctions can disrupt the stability of the system and even lead to task failure. Therefore, this application models a multi-robot system, wherein the first... Dynamics model of a robot Represented as: ,in, , Indicates the total number of multi-robot systems; Indicates the first A robot's state vector contains measurable dynamic information such as its position and velocity, which can usually be obtained through sensors; These are the input signals designed to control the robot, i.e., the controller output; This indicates an actuator malfunction (such as a motor or servo motor) that affects the actuator itself, such as a decrease in torque, jamming, or complete failure. It is the system output information measured by sensors that reflects the actual operating status of the robot. ; This indicates a sensor malfunction, such as measurement drift, sudden noise increase, or data loss; matrix These are constant matrices describing the dynamic characteristics of the robot system, representing system state transition, control input gain, output observation, actuator failure effects, and sensor failure effects, respectively. This model explicitly incorporates actuator and sensor failures into the system dynamics, providing a theoretical foundation for the subsequent design of fault-tolerant observers and distributed cooperative controllers.

[0075] In some embodiments of this application, in order to design a distributed cooperative fault-tolerant controller based on an intermediate observer, the original robot model is state-extended to uniformly handle system state and actuator fault information. The dynamic model of the robot is extended to , specifically, , ,in, This represents the expanded state vector, derived from the original system state. and actuator fault estimation composition; It is the state transition matrix of the extended state system, which keeps the original system dynamics unchanged while allowing the fault terms to evolve independently; The matrix represents the effect of control inputs on the extended state, affecting only the original system state. It is the effect matrix of actuator failure on the system and the fault estimation state. Its structure ensures that the fault can simultaneously affect the system state and its own estimation terms. It is the influence matrix of sensor faults on extended states, indicating that sensor faults only affect the output and do not directly change the state or fault estimate; This is an extended output matrix that maps both the system output and sensor faults to the observed output, allowing for indirect identification of sensor faults. This state extension method achieves joint modeling of system states and actuator faults, providing a unified mathematical framework for the subsequent design of intermediate observers.

[0076] In some embodiments of this application, in practical application scenarios, actuator and sensor faults are often unknown and time-varying, making them difficult to measure directly. To achieve simultaneous estimation of system state and fault state, this application introduces a quantitative relationship between actuator faults and extended system states, and designs a fault estimator based on intermediate variables. First, actuator faults are defined. With system extended state The relationship between them can be represented as: ,in, It is an auxiliary variable representing the deviation between actuator failure and its dynamic coupling term in the system; It is a positive scalar gain used to adjust the strength of the coupling relationship; It reflects the direction of the fault's effect on the system; this formula shows that the actuator fault can be modeled as a linear function of the system state plus an unknown disturbance, thus providing a theoretical basis for the subsequent construction of intermediate variables.

[0077] Based on the above relationships, this application further designs an intermediate observer that satisfies the following representation: ,in, This represents an estimate of the extended state of the system, including the original state and actuator failure estimates. This is an estimate of the actuator malfunction; The estimation error of the robot relative to its output information, which incorporates the deviation between the neighbor's estimate and the leader's actual output, satisfies the following formula: ; It is the adjacency weight in the communication graph, representing the robot's... with neighbors The connection strength; It is a positive scalar gain, which controls the strength of the feedback term. It is the designed gain matrix, used to adjust the convergence speed and stability of the observer.

[0078] In some embodiments of this application, in order to achieve joint estimation of actuator faults and system states, this application introduces an intermediate variable based on the extended state model. This is used to characterize the deviation between actuator failure and system dynamics. Based on this, the dynamic equation of the intermediate variable estimator is designed as follows: ; in, It is an intermediate variable The estimated value is dynamically driven by the system state, control input, and output error. It is the designed gain matrix used to adjust the weight of error feedback; the dynamic equation achieves asymptotic reconstruction of unknown fault terms by combining the estimation error of intermediate variables with the system input, state and output feedback, without relying on prior knowledge of the fault amplitude or its rate of change.

[0079] In some embodiments of this application, to perform subsequent stability analysis and performance verification of the fault estimator and state estimator, this application further constructs a robot system state estimation error model and an intermediate variable estimation error model. By merging the two to form a new composite state variable, the design process of the collaborative fault-tolerant controller is simplified. First, the dynamic model of the estimation error of the state estimator is defined as follows: ; in, This represents the estimation error of the extended state, that is, the deviation between the actual extended state and the estimated value; It is the estimation error of the intermediate variable; It represents the estimation error of sensor faults; this model describes the evolution of the state estimation error under the action of the observer, reflecting the influence of system dynamics, fault terms, and feedback correction.

[0080] Secondly, the estimation error model for the intermediate variables is designed as follows: ; in, It represents a real actuator failure. The model characterizes the dynamic behavior of the intermediate variable estimation error, and its structure ensures that the estimation error asymptotically approaches zero when there is no failure or the failure is constant.

[0081] Through the two error models mentioned above, this application achieves unified modeling of the estimation errors of state and intermediate variables, providing a theoretical basis for subsequent controller design.

[0082] In some embodiments of this application, in order to enable unified analysis of the state and fault state of the robot system, this application constructs a new composite state variable. This variable will estimate the error in the extended state of the system. estimation error of intermediate variables This is integrated into a single overall state vector, thereby achieving joint modeling of the state estimation and fault estimation errors. Furthermore, this application merges the state estimation error model with the intermediate variable estimation error model, constructing its extended dynamic model as follows: ; in, It is a matrix of the original system and The resulting constant augmented matrix has the following specific form: This matrix describes the internal dynamic evolution relationship of the composite state error; It is the feedback gain matrix, which is composed of the gains corresponding to the state error and the intermediate variable error; It is an external disturbance input matrix that reflects the impact of sensor failures on the system; It includes actual actuator failures. and sensor failure The perturbation vector. This extended model achieves a unified expression for the estimation errors of the state and intermediate variables, providing a concise and consistent mathematical framework for subsequent controller design.

[0083] Therefore, this application achieves joint estimation of system state, actuator faults, and sensor faults by introducing an intermediate variable, thereby overcoming the dependence of traditional methods on prior knowledge and structural constraints in fault estimation. Specifically, this method first constructs an intermediate variable, which consists of a linear combination between system state and actuator faults. This embeds fault information that was originally not directly observable into the observer dynamics, transforming it into a reconfigurable auxiliary dynamic. This avoids pre-assumptions about fault amplitude, rate of change, or specific form, achieving estimation capabilities "independent of prior fault knowledge."

[0084] Based on this, by designing an intermediate observer, the system input-output relationship and historical data trends are used to generate joint estimates of system state, actuator faults, sensor faults and intermediate variables in real time. Sensor faults are indirectly estimated through an extended state model. All estimates together constitute a "joint estimator" for subsequent control decisions.

[0085] This scheme further decouples the strong constraints between faults and the system output matrix, breaking the traditional observer design limitation that requires the "fault gain matrix to output matrix matching condition." This makes the observer structure no longer dependent on a specific system topology or output configuration, applicable to a wider range of system structures. Simultaneously, by introducing a bias feedback mechanism—that is, using relative output error to drive estimation error convergence—the observer's adaptability to dynamic disturbances is enhanced, improving estimation accuracy and robustness. Ultimately, this technical solution achieves high-precision, high-reliability collaborative fault-tolerant control of multi-robot systems without requiring precise fault models or strict observer matching conditions, significantly improving the system's applicability and stability in complex real-world environments.

[0086] In some embodiments of this application, after establishing the state estimation error and intermediate variable estimation error models, a cooperative error model for a multi-robot system is further constructed to design a distributed cooperative fault-tolerant controller. This model combines the state information of the follower robot and the leader robot, and utilizes the properties of the Laplace matrix to describe the information interaction relationship under the network topology. The designed robot cooperative error model is as follows: ; in, It is the first The collaborative error state of a follower robot represents the deviation between its own state and the expected behavior of the group. It is a composite feedback matrix composed of state- and fault-related gains; where, ,in It is a positive scalar, used to adjust the strength of state feedback; This is used to compensate for the effects of actuator failure; It is an estimate of the composite state error variable, which includes the estimation errors of the state and intermediate variables; It is a collaborative control gain designed based on actual application scenarios, used to adjust the tracking speed of neighbors and leaders; , These are the state estimates of the neighbor and the local machine, respectively. It is the weight of communication with the leader, representing the first The ability of a follower to obtain status information from the leader; This is the actual state of the leader robot; These are extended state terms that reflect the dynamic coupling of the system; This is the direct disturbance term of actuator failure to the coordination error. This model comprehensively considers state estimation error, network topology, fault effects, and coordination control requirements, providing a complete mathematical foundation for subsequent controller design.

[0087] In some embodiments of this application, the above technical solution process is for model building of a single robot. To realize the model building of multiple robots, it is necessary to use the knowledge of Kronecker product in matrix theory to transform the state estimation error and intermediate variable estimation error of a single robot, as well as the cooperative error model, into the Kronecker product form of multiple robots.

[0088] In some embodiments of this application, to ensure stable cooperative control of the multi-robot system while effectively avoiding adverse effects of actuator and sensor failures on the system state, this application designs a distributed cooperative fault-tolerant controller based on state estimation and intermediate variable estimation information. The specific form of this controller is as follows: ; in, It is the first Action control commands for a follower robot; composite state estimation vector This controller achieves robust cooperative control in the presence of actuator and sensor failures by fusing state estimation, fault estimation, and network topology information, ensuring that the system can stably track the leader's behavior.

[0089] In some embodiments of this application, since different types of robots have different driving methods and degrees of freedom structures, the physical meaning of their control inputs differs fundamentally. Therefore, before generating motion control instructions, the distributed cooperative fault-tolerant controller determines the specific form of the motion control instructions based on a preset cooperative control objective and the robot platform type, including but not limited to: (1) For wheeled mobile robots, motion control commands include linear velocity and angular velocity. Thus, using linear velocity and angular velocity as direct control quantities facilitates planar motion tracking.

[0090] (2) For legged robots or robotic arms, motion control commands include joint angles and joint torques to precisely regulate limb posture and power output.

[0091] (3) For UAVs, motion control commands include pitch angle, roll angle, yaw angle and throttle control to coordinate flight attitude and altitude.

[0092] This adaptation process ensures that the generated control commands not only meet the requirements of the collaborative task but also match the execution capabilities of each platform, thereby guaranteeing the effective implementation of fault-tolerant control strategies in heterogeneous multi-robot systems.

[0093] In some embodiments of this application, the method further includes: The number of follower robots and the type of communication topology are determined based on the actual task scenario. When the number of follower robots is less than a preset threshold and the communication environment is good, an undirected graph topology is used to ensure the symmetry and integrity of information interaction. When the number of follower robots exceeds the preset threshold or the communication bandwidth is limited, a directed graph topology is used to reduce network load.

[0094] Specifically, the scale and communication structure of the multi-robot system are rationally configured based on actual task requirements and operating environment conditions. By analyzing the number of follower robots and the quality of the communication environment, a suitable communication topology is dynamically selected to achieve a balance between system performance and resource consumption. When the number of robots is small (e.g., less than 10) and communication conditions are good, an undirected graph topology ensures bidirectional information exchange between any two connected robots, thereby guaranteeing the symmetry and integrity of information interaction and improving collaborative accuracy and consistency. However, when the number of robots is large (e.g., more than 10) or communication bandwidth is limited, a directed graph topology is adopted, maintaining only the necessary unidirectional information flow, effectively reducing communication overhead and network congestion, lowering the overall network load, and enhancing the scalability and stability of the system in large-scale or resource-constrained scenarios. This design allows the system to flexibly adapt to different task scales and communication conditions, providing a suitable network foundation for subsequent fault-tolerant control.

[0095] In some embodiments of this application, each robot obtains its own status information through an external positioning system: a global positioning system is used in outdoor environments, and an ultra-wideband, infrared, or visual motion capture system is used in indoor environments. Simultaneously, it periodically exchanges position, speed, orientation, and remaining battery power information with neighboring robots via a wireless communication module.

[0096] Specifically, in different operating environments, by adapting appropriate external positioning technologies, the robots can continuously obtain high-precision position and attitude data: in open outdoor scenarios, global positioning is achieved using the Global Positioning System (GPS); in complex indoor environments or environments with many obstructions, high-precision indoor positioning methods such as ultra-wideband, infrared, or visual motion capture systems are employed. Simultaneously, each robot periodically broadcasts its status parameters, including position, speed, orientation, and remaining battery power, to its neighbors via wireless communication modules. This periodic information exchange not only provides the necessary external input for collaborative control and fault-tolerant estimation but also enhances the system's global perception of individual states, laying the foundation for stable and efficient distributed collaboration.

[0097] In some embodiments of this application, the method further includes: each follower robot acquiring its own state output information through an inertial measurement unit, lidar, vision sensor, or encoder. Under normal operating conditions, the robot inputs its own state output information as actual output information to an intermediate observer. When a sensor failure occurs, the intermediate observer estimates the type, magnitude, and occurrence time of the sensor failure in real time based on the input-output relationship of the dynamic model and historical data trends. It then fuses the estimated sensor failure with partially reliable self-state output information to generate a compensated system state estimate, which is used as part of the joint estimate by the distributed cooperative fault-tolerant controller. This ensures that even with partial sensor failure, the system can still obtain accurate and continuous state feedback, maintaining the stability and effectiveness of cooperative control.

[0098] In some embodiments of this application, the method further includes: performing a real-time assessment of the health status of the actuator before the actual execution of the control command, to ensure the safe and effective implementation of the motion command. Specifically, before applying the motion control command to the actuator of the corresponding follower robot, based on the actuator's current, voltage, output torque, and response speed parameters, combined with preset health status thresholds and dynamic performance indicators, it is determined whether the actuator has malfunctioned.

[0099] If no fault occurs, the action control command is executed directly, and the execution status is fed back to the intermediate observer to update the system output information and maintain the estimation accuracy.

[0100] If an actuator failure is detected, a fault tolerance mechanism is triggered. The intermediate observer estimates the type, extent, and executable capability of the actuator failure in real time based on the actuator's dynamic model and actual output data. The estimated actuator failure is then used to correct the intermediate variable estimates, thereby compensating for the actuator failure when generating subsequent motion control commands. In other words, the actual capability limitations of the actuator are proactively considered, and dynamic compensation for the actuator failure is achieved, ensuring that the system can still operate stably and complete collaborative tasks even with partial loss of execution capability.

[0101] Secondly, refer to Figure 2 This application provides a computer device that is deployed in each robot of a multi-robot system. The computer device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned distributed fault-tolerant control method for the multi-robot system.

[0102] In summary, the distributed fault-tolerant control method and apparatus for multi-robot systems provided in this application have the following technical effects.

[0103] This method constructs a unified dynamic model incorporating actuator and sensor faults, introduces intermediate variables relating to system state and actuator faults, and designs an intermediate observer that is independent of prior knowledge of fault amplitude and its rate of change, and does not require the system to meet observer matching conditions. This enables joint real-time estimation of system state, actuator faults, sensor faults, and intermediate variables. Based on this, a distributed cooperative fault-tolerant controller is constructed by combining communication topology and external state information. This controller generates motion control commands that effectively compensate for the impact of faults, ensuring that all follower robots can stably and consistently track the leader robot's motion even in the presence of actuator or sensor faults. The entire scheme exhibits strong robustness, high adaptability, and good engineering deployability, significantly improving the cooperative reliability and task continuity of multi-robot systems in complex and uncertain environments.

[0104] It should be noted that in all specific embodiments of this application, all data processing activities related to user identity or personal characteristics, such as user information, user behavior data, historical data, and location information, will be conducted in accordance with the principles of legality, legitimacy, and necessity. All data collection, use, storage, and processing will be subject to compliance with applicable national and regional laws, regulations, and industry standards, and informed consent from users will be obtained in a clear and explicit manner before processing. For the processing of sensitive personal information, separate consent from users will be obtained through prominent means such as pop-up prompts and independent confirmation pages. If any processing conflicts with laws and regulations, the laws and regulations will prevail, and necessary data processing will only be carried out within the scope permitted by laws and regulations, ensuring that all data-based applications, analyses, and technical implementations are conducted within the scope permitted by laws and regulations.

[0105] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0106] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of ordinary skill of an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary skill. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0107] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several programs to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0108] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable programs for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can retrieve and execute a program from or in conjunction with such a program execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with a program execution system, apparatus, or device.

[0109] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Additionally, computer-readable media can even be paper or other suitable media on which programs can be printed, for example, by optically scanning the paper or other media, then editing, interpreting, or, if necessary, processing it in a suitable manner to obtain the program electronically, and then storing it in computer memory.

[0110] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable program execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0111] In the foregoing description of this specification, the reference to terms such as "one embodiment / implementation," "another embodiment / implementation," or "certain embodiments / implementations," etc., indicates that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in an embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0112] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0113] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of the present invention.

Claims

1. A distributed fault-tolerant control method for a multi-robot system, characterized in that, The multi-robot system includes a leader robot and multiple follower robots. Each follower robot obtains the state information of the leader robot and neighboring robots as external state information through a communication network. The method includes the following steps: Construct a dynamic model of the multi-robot system, including actuator failures and sensor failures; For each follower robot, an intermediate observer is constructed by introducing intermediate variables to correlate system state and actuator failure, combined with the output information of the dynamic model; The intermediate observer generates joint estimates of system state, actuator faults, sensor faults, and intermediate variables in real time. Based on the joint estimate and external state information, and combined with the communication topology, a distributed collaborative fault-tolerant controller is constructed to generate action control commands; The motion control commands are applied to the actuators of the corresponding follower robots, enabling all follower robots to consistently track the movement state of the leader robot.

2. The distributed fault-tolerant control method for a multi-robot system according to claim 1, characterized in that, The construction of the dynamic model of the multi-robot system includes the following steps: Each robot's motion state is described as a system state including its position and velocity; Actuator failures are modeled as unknown abnormal deviations acting on the robot control input channel, which are used to characterize actual failure scenarios such as actuator performance degradation, jamming, or failure. Sensor faults are modeled as unknown disturbances superimposed on the robot's output signal to reflect actual measurement anomalies such as sudden increases in measurement noise, sensor drift, or data loss. Based on the dynamic correlation between the system state, control input, actuator faults, and sensor faults, a dynamic model reflecting the dynamic characteristics of robot motion is established.

3. The distributed fault-tolerant control method for a multi-robot system according to claim 1, characterized in that, For each follower robot, an intermediate observer is constructed by introducing intermediate variables to correlate system state and actuator failure, combined with the output information of the dynamic model, including the following steps: An intermediate variable is constructed for each follower robot, and the intermediate variable is determined by a linear combination of actuator failure and system state; Based on the output information of the intermediate variables and the dynamic model, the local observer dynamic equation is established to form an intermediate observer structure suitable for a single follower robot. The local intermediate observers of each follower robot are interconnected through a communication network to form a distributed intermediate observer.

4. The distributed fault-tolerant control method for a multi-robot system according to claim 1, characterized in that, The process of generating joint estimates of system state, actuator faults, sensor faults, and intermediate variables in real time through the intermediate observer includes the following steps: The intermediate observer generates estimated output information for each follower robot based on its internal dynamic structure. The deviation between the actual output information and the estimated output information of the follower robot is calculated, and the deviation is input into the dynamic equation of the intermediate observer to update its internal state; Based on the updated internal state of the intermediate observer, the system state, actuator faults, sensor faults, and estimated values ​​of intermediate variables are synchronously output to form the joint estimate.

5. The distributed fault-tolerant control method for a multi-robot system according to claim 1, characterized in that, The process of constructing a distributed collaborative fault-tolerant controller and generating action control commands based on the joint estimate and external state information, combined with the communication topology, includes the following steps: Obtain the system state estimate and intermediate variable estimate for each follower robot; the intermediate variable estimate is used to equivalently characterize actuator failure. Receive external status information from neighboring robots and the leader robot; The information connection relationship between each follower robot and its neighbors is determined based on the communication topology. Based on the system state estimate, the external state information, and the communication topology, the state deviation of each follower robot relative to its neighboring robots and the leader robot is calculated as the cooperative error. The collaborative error is combined with the estimated value of the intermediate variable, and the combined result is subjected to amplitude limiting and smoothing to generate motion control commands.

6. The distributed fault-tolerant control method for a multi-robot system according to claim 5, characterized in that, Before generating the motion control command, the distributed collaborative fault-tolerant controller determines the specific form of the motion control command based on the preset collaborative control objective and the robot platform type: for wheeled mobile robots, the motion control command includes linear velocity and angular velocity; for legged robots or robotic arms, the motion control command includes joint angle and joint torque; for drones, the motion control command includes pitch angle, roll angle, yaw angle and throttle control amount.

7. The distributed fault-tolerant control method for a multi-robot system according to claim 1, characterized in that, The method further includes: The number of follower robots and the type of communication topology are determined based on the actual task scenario. When the number of follower robots is less than a preset threshold and the communication environment is good, an undirected graph topology is used to ensure the symmetry and integrity of information interaction. When the number of follower robots exceeds a preset threshold or the communication bandwidth is limited, a directed graph topology is used to reduce network load. Each robot obtains its own status information through an external positioning system: a global satellite positioning system is used in outdoor environments, and an ultra-wideband, infrared, or visual motion capture system is used in indoor environments; at the same time, it periodically exchanges position, speed, orientation, and remaining battery information with neighboring robots through a wireless communication module.

8. The distributed fault-tolerant control method for a multi-robot system according to claim 1, characterized in that, The method further includes: each follower robot acquiring its own state output information through an inertial measurement unit, lidar, vision sensor or encoder; Under normal operating conditions, the self-state output information is input to the intermediate observer as the actual output information. When a sensor fault occurs, the intermediate observer estimates the type, magnitude, and occurrence time of the sensor fault in real time based on the input-output relationship of the dynamic model and historical data trends. The estimated sensor fault is then fused with some reliable self-state output information to generate a compensated system state estimate, which is used as part of the joint estimate by the distributed collaborative fault-tolerant controller.

9. The distributed fault-tolerant control method for a multi-robot system according to claim 1, characterized in that, The method further includes: before applying the motion control command to the actuator of the corresponding follower robot, determining whether the actuator has malfunctioned based on the actuator's current, voltage, output torque, and response speed parameters, combined with a preset health status threshold and dynamic performance indicators; if no malfunction occurs, the motion control command is executed directly, and the execution status is fed back to the intermediate observer to update the system output information; if an actuator malfunction is detected, the intermediate observer estimates the type, degree, and executable capability of the actuator malfunction in real time based on the actuator's dynamic model and actual output data, and uses the estimated actuator malfunction to correct the intermediate variable estimate, thereby compensating for the actuator malfunction when generating subsequent motion control commands.

10. A computer device, characterized in that, The computer device is deployed in each robot of the multi-robot system. The computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the distributed fault-tolerant control method for the multi-robot system according to any one of claims 1 to 9.

Images