A data-model hybrid driven air-ground fixed-time fast fault-tolerant control method

Abstract

The application discloses a data-model hybrid driving air-ground fixed time fast fault-tolerant control method, specifies a ground unmanned vehicle as a leader, the rest as followers, constructs a leader-following configuration and an expected task, uses weighted graph theory, a Laplacian matrix and a signed graph to depict topological interaction, competitive relationship and interest demand influence; interference and actuator fault influence are analyzed, a fault model is established, a distributed error system is linearized and parameters are identified online, and the fault model is formed; a distributed data-driven fixed time observer is designed for each follower, the leader state is estimated, and a communication loop is eliminated; a preset performance function is used to convert an error constraint problem, adaptive and optimal control are combined, a data-model hybrid driving fault-tolerant controller is designed, and low energy consumption and fixed time task completion under faults are ensured. The application does not depend on modeling accuracy to realize direct compensation for complex uncertainty, and has better fault-tolerant performance.

Description

Technical Field

[0001] This invention relates to the field of consistency-based rapid fault-tolerant control of air-ground cross-domain collaborative systems composed of multiple unmanned helicopters and multiple ground unmanned vehicles, and more specifically to a data-model hybrid driven air-ground fixed-time rapid fault-tolerant control method. Background Technology

[0002] Cross-domain unmanned collaboration refers to the organic integration of multiple unmanned systems with significant functional differences that can operate in different spatial domains, such as land, sea, air, and space. These systems can complement each other's functions and multiply their energy efficiency through information sharing and fusion, behavioral interaction and coordination, and task collaboration and cooperation, thereby enhancing their ability to cope with complex environments and missions. Low-altitude unmanned helicopters possess a top-down perspective and wide-area coverage capabilities, while ground-based unmanned vehicles can carry various high-precision sensors and have good near-field recognition capabilities. Therefore, unmanned helicopter-ground unmanned vehicle collaborative perception is beneficial for fully leveraging the functional redundancy and complementary capabilities of heterogeneous platforms, improving the overall perception accuracy and robustness of the system. In recent years, with the continuous deepening of research on air-ground integrated systems, related technologies have evolved from multi-body collaboration to air-ground cross-domain collaboration. Constructing an air-ground cross-domain collaborative control system with high spatiotemporal coordination and stability is of great significance for ensuring the autonomy and safety of unmanned systems in engineering missions.

[0003] However, external disturbances and model uncertainties in complex environments further exacerbate the dynamic nonlinearity of the system, making it difficult for traditional linearized control methods to meet increasingly stringent requirements for accuracy and real-time performance. Furthermore, mechanical fatigue and component wear caused by prolonged operation, resulting in actuator and sensor failures, are common threats to system safety. In multi-agent formation systems, individual failures or attacks on the communication network can dynamically alter the size and topology of the system, affecting the overall performance of the formation and potentially leading to mission failure or even major air disasters. Therefore, achieving rapid, coordinated, fault-tolerant control of air-to-ground systems under heterogeneous dynamics and sudden failure conditions has become a critical scientific problem urgently needing to be solved. While some research has been conducted on the fault-tolerant control of heterogeneous multi-agent systems, the question of how to construct a fixed-time control strategy for air-ground cross-domain systems based on non-equilibrium interaction of multi-source heterogeneous digital information, rule-based experience knowledge, and advanced control technologies in complex and harsh environments, and to achieve data-knowledge hybrid-driven adaptive fault-tolerant control, remains unsolved. Therefore, it is necessary to conduct research on new safe fault-tolerant control technologies to ensure that air-ground cooperative systems possess new operational advantages such as rapid environmental adaptation, safe fault-tolerant control, and autonomous performance optimization in complex environments. Summary of the Invention

[0004] Purpose of the invention: To address the above-mentioned problems, this invention proposes a data-model hybrid driven air-to-ground fixed-time fast fault-tolerant control method, which can ensure that the system can complete the expected tracking task within a fixed time within a pre-designed time period, while simultaneously constraining the transient and steady-state performance of the system.

[0005] Technical Solution: The present invention discloses a data-model hybrid driven air-to-ground fixed-time fast fault-tolerant control method, applied to a cross-domain heterogeneous multi-agent system composed of multiple unmanned helicopters and multiple ground unmanned vehicles, comprising the following steps:

[0006] (1) Determine that the unmanned vehicle on the road is the leader and the other unmanned vehicles are the followers, and clarify the collaborative configuration and expected tasks of the multi-agent system; use weighted graph theory, symbolic graph and Laplace matrix to quantify the topological changes brought about by information interaction, competitive and cooperative relationship and differences in interests among agents;

[0007] (2) Analyze the impact of external disturbances and actuator failures on the system and establish a fault description model based on prior knowledge; then, for the distributed error system of the multi-agent system, establish a distributed compact linearization model with unknown parameters by means of the equivalent dynamic linearization method, and use the interaction data between agents to identify and optimize the unknown parameters of the linear model online, and finally form a fault model of the heterogeneous multi-agent system that can describe the main nonlinear complex characteristics.

[0008] (3) To address the problem that followers cannot obtain the time-varying state information of the leader, based on distributed theory and fixed-time theory, a distributed data-driven fixed-time observer under a non-fully interactive topology network is designed for each follower to achieve effective estimation of the leader's unknown state information, eliminate the communication loop problem existing in each follower, and provide support for the design of subsequent control methods and the formation of a reference tracking trajectory for global followers.

[0009] (4) Construct a fixed-time preset performance function and transform the tracking problem with error constraints into an unconstrained stabilization problem through a specific error transformation; combine adaptive fixed-time control technology and optimal control technology to construct a data-model hybrid driven fixed-time fault-tolerant controller for heterogeneous multi-agent systems, so as to realize the discrete-time multi-agent system under actuator failure to reproduce complex expected tasks in a fixed time with low energy consumption.

[0010] Furthermore, the implementation process of step (1) is as follows:

[0011] Based on graph theory, a directed graph with N followers is used for communication. To describe, among which, and These represent the set of heterogeneous multi-agent systems, the set of communication connection edges, and the connection weight matrix of the communication topology network, respectively. as well as These represent the i-th agent being unable to communicate with the j-th agent and being able to communicate with the i-th agent, respectively. and These represent the i-th follower and the j-th follower being in a friendly cooperative relationship and a competitive relationship, respectively; when At this time, the communication network is a special type of undirected graph network, where the agents communicating with each other have the same weight; the in-degree matrix of the network is represented as... ,in, The Laplace matrix of a communication network is represented as follows: To describe the interaction between leaders and followers, a matrix is ​​designed. To indicate, among which For the connection weight between follower i and leader, when When , it means that the i-th follower can obtain the leader's information, otherwise they cannot.

[0012] Furthermore, the implementation process of step (2) is as follows:

[0013] Cross-domain heterogeneous multi-agent systems include an autonomous vehicle leader and There are several followers, including unmanned helicopters and unmanned ground vehicles. To construct a leader-follower consistency controller, the following discrete-time motion model of the leader is constructed:

[0014]

[0015] in, and These represent the system state and control input of the leader at time k, respectively. and A system matrix with appropriate dimensions;

[0016] Secondly, the first The dynamic model of a single follower simplifies to the following discrete-time expression:

[0017]

[0018] in, Let K represent the follower's input state and control input data at time k, respectively. Represents an unknown positive definite constant. This represents an unknown function describing the nonlinear characteristics of a follower system;

[0019] To describe the characteristics of actuator failure, design Using the actual control input and the desired input signal to represent these, we obtain the actuator fault model shown below:

[0020]

[0021] in, Denotes an unknown efficiency loss factor, satisfying ;

[0022] Taking actuator failure into account, the first The dynamic model of a follower, i.e., the fault description model based on prior knowledge, is further expressed as:

[0023]

[0024] To facilitate controller design, the following distributed consistency tracking error is constructed:

[0025]

[0026] in, Represents a symbolic function. This represents the leader's trajectory, i.e., the desired tracking signal. To fully utilize control input and tracking error data, a dynamic linearization technique is employed to construct the following distributed compact-form linearization model:

[0027]

[0028] in, , This represents the time-varying pseudopartial derivative parameter.

[0029] Furthermore, the distributed data-driven fixed-time observer in the non-fully interactive topology network described in step (3) is:

[0030]

[0031] in, This represents the weight of the communication connection between two agents. Indicates the state of the leader The estimated value; the observer parameters satisfy: ,also, All are positive definite parameters to be designed; when the parameter matrix satisfies:

[0032]

[0033] in, It is a positive definite constant. yes 3D identity matrix It is a 1-dimensional identity matrix. For a system matrix of appropriate dimension, If we consider the extended Laplace matrix of the communication network, then the follower agent will acquire the state information of the leader agent within a fixed time interval.

[0034] Furthermore, the implementation process of step (4) is as follows:

[0035] Design the tracking error variables for the following multi-agent system:

[0036]

[0037] To achieve performance constraints, a new controlled variable is constructed using a fixed-time pre-set performance method:

[0038]

[0039] in, It is the default performance function, and its expression is:

[0040]

[0041] in, All are positive definite constants. Indicates the initial value. This indicates the preset fixed convergence time;

[0042] Based on the fixed-time theorem and a preset performance control strategy, a data-driven fault-tolerant linearized controller is constructed using a linearized model and system input / output data, such that the following equation holds:

[0043]

[0044] To design a controller with pre-defined ideal performance for completing a consistent bounded tracking task within a fixed time, its construction is as follows:

[0045]

[0046] in, Represents an unknown nonlinear function of the model. This indicates the convergence error of the actual error with respect to the expected value. Assuming an ideal error value, construct the following control input:

[0047]

[0048] in, It is an unknown control gain, and:

[0049]

[0050] Based on the above control inputs, the global error achieves fixed-time convergence, and the performance indicators meet the constraints of the preset performance function.

[0051] The following design incorporates an adaptive law for the controller parameters: First, construct a law regarding... Cost function:

[0052]

[0053] in:

[0054]

[0055] And design the model error compensator as shown below:

[0056]

[0057] Based on optimal control theory, its bias is calculated. ,get:

[0058]

[0059] Furthermore, we have:

[0060]

[0061] as well as:

[0062]

[0063] The following parameters were obtained. Adaptive regulation law:

[0064]

[0065] And satisfy:

[0066]

[0067] Design about parameters Cost function:

[0068]

[0069] Its adaptive regulation law is derived as follows:

[0070]

[0071] And satisfy:

[0072] .

[0073] Beneficial Effects: Compared with existing technologies, the present invention offers the following advantages: Based on sampled data, the present invention designs a fixed-time leader observer for each follower, solving the problem of specified-time observation in discrete-time multi-agent systems; the novel fixed-time performance function designed in this invention can impose preset transient and steady-state performance constraints on tracking errors in the discrete-time domain; utilizing the equivalent dynamic linearization method, combined with fixed-time preset performance and fixed-time observations, a novel fixed-time model-free controller is constructed, including an estimation criterion for a model compensator, an adaptive adjustment law for optimizing pseudo-partial derivative parameters, and an adaptive model-free data-driven optimal fault-tolerant controller obtained using gradient descent; the constructed model error compensator achieves direct compensation for complex uncertainties, independent of modeling accuracy; and possesses superior fault-tolerant performance, capable of solving the fault regulation problem of unknown multi-agent systems in the presence of external disturbances, model uncertainties, and actuator failures. Attached Figure Description

[0074] Figure 1 This is a flowchart of the present invention;

[0075] Figure 2 This is a schematic diagram of the leader-follower formation system;

[0076] Figure 3 This is a communication topology diagram of the leader-follower formation system;

[0077] Figure 4 It is a tracking graph of the followers' positions relative to the leader on the X-axis;

[0078] Figure 5 It is a tracking graph of the followers' positions relative to the leader on the Y-axis;

[0079] Figure 6 It is a tracking graph of the follower's position relative to the leader on the Z-axis;

[0080] Figure 7 It is the convergence plot of the consensus error of heterogeneous multi-agents along the X-axis;

[0081] Figure 8 It is the convergence plot of the consensus error of heterogeneous multi-agents along the Y-axis;

[0082] Figure 9 It is the convergence plot of the consistency error of heterogeneous multi-agents along the Z-axis. Detailed Implementation

[0083] The invention will now be further described with reference to the accompanying drawings.

[0084] like Figure 1As shown, this invention proposes a data-model hybrid driven, air-to-ground fixed-time fast fault-tolerant control method. It designs a communication topology network weight model for a multi-agent system, employs a dynamic linearization method to establish an equivalent linearized model for a data-driven heterogeneous formation system, and uses adaptive pseudo-partial derivative parameters to improve modeling accuracy. For discrete-time multi-agent systems, in the presence of actuator failures and unknown uncertainties, a fixed-time observer is designed for each follower agent to estimate the information required by the leader agent. This observer can complete its observation task within a preset fixed time. The constrained error variable is transformed using a strictly incremental preset performance function. An adaptive fixed-time fault-tolerant tracking controller is designed using the transformed error and only the input-output data of the discrete-time multi-agent system. The specific implementation process is as follows:

[0085] Step 1: As Figure 2 As shown, an ideal unmanned ground vehicle is designed as the leader, and other unmanned vehicles are followers. A global leader-follower consistency configuration and expected tasks are constructed. For topological networks with dynamic and unbalanced information interaction, weighted graph theory is used to describe the information interaction capabilities of different agents. Then, a global Laplace matrix is ​​constructed to describe the interaction topology of the entire heterogeneous multi-agent system, and the impact of differences in interests on the topology is analyzed. Symbolic graphs are used to characterize the competition and cooperation relationships between agents.

[0086] This invention studies the complex information interaction between unmanned vehicles (UAVs) in a cross-domain collaborative system, involving competition and cooperation via a communication network. A directed graph with different weights is designed to describe the interaction. Based on graph theory, a directed graph with N followers can be used... To describe, among which, and These represent the set of heterogeneous multi-agent systems, the set of communication connection edges, and the connection weight matrix of the communication topology network, respectively. as well as These represent the i-th agent being unable to communicate with the j-th agent and being able to communicate with the i-th agent, respectively. and Let represent the i-th follower and the j-th follower, respectively, their relationships as friendly cooperation and competition. At this point, the communication network is a special type of undirected graph network, meaning that the agents communicating with each other have the same weight. The in-degree matrix of the network is represented as... ,in, Then, the Laplace matrix of the communication network can be represented as... To describe the interaction between leaders and followers, a matrix was designed. To indicate, among which For the connection weight between follower i and leader, when When , it means that the i-th follower can obtain the leader's information, otherwise they cannot.

[0087] It is worth noting that in existing research, agents mostly connect with their neighbors in a cooperative manner. However, in many real-world scenarios, the relationships between agents may be competitive-cooperative or purely competitive, similar to friendly or hostile relationships in human society. Therefore, this invention utilizes graph theory to construct a topological description model of hybrid interaction relationships, which has significant practical implications for studying the consistency of multi-agent systems in hybrid interaction relationships and tasks such as formation tracking.

[0088] Step 2: Accurately grasp the state evolution relationship of the multi-agent system under external disturbances and actuator failures, as well as the impact mechanism of failures on system consistency and tracking performance, and establish a fault description model based on prior knowledge. Then, for the distributed error system of the multi-agent system, a distributed compact linearized model with unknown parameters is established using the equivalent dynamic linearization method. The unknown parameters of the linear model are identified and optimized online using the interaction data between agents, ultimately forming a fault model of the heterogeneous multi-agent system that can describe the main nonlinear complex characteristics.

[0089] The unmanned helicopter-ground unmanned vehicle cross-domain system includes an unmanned vehicle leader and There are several followers, including unmanned helicopters and unmanned ground vehicles. To construct a leader-follower consistency controller, the following discrete-time motion model of the leader is built:

[0090]

[0091] in, and These represent the system state and control input of the leader at time k, respectively. and It is a system matrix with appropriate dimensions.

[0092] Secondly, the first The dynamic model of a single follower simplifies to the following discrete-time expression:

[0093]

[0094] in, Let K represent the follower's input state and control input data at time k, respectively. Represents an unknown positive definite constant. This represents an unknown function describing the nonlinear characteristics of a follower system. In this invention, the dynamic characteristics and input signals differ among different individuals.

[0095] Considering the complex structure and high operational load of intelligent agents, coupled with their potential strong coupling and dynamic instability, actuators are prone to failure. To describe the characteristics of actuator failures, a design is developed... Let these represent the actual control input and the desired input signal. Based on this, the actuator fault model can be obtained as follows:

[0096]

[0097] in, Denotes an unknown efficiency loss factor, satisfying .

[0098] Taking actuator failure into account, the first The dynamic model of a single follower can be further represented as:

[0099]

[0100] To facilitate controller design, the following distributed consistency tracking error is constructed:

[0101]

[0102] in, Represents a symbolic function. This represents the leader's trajectory, i.e., the desired tracking signal. To fully utilize control input and tracking error data, this invention employs dynamic linearization technology to construct the linearization model shown below:

[0103]

[0104] in, , The time-varying pseudopartial derivative parameter is represented by this parameter. This model does not rely on prior knowledge of the nonlinear model of the multi-agent system, greatly simplifying the modeling process. It eliminates the need for system parameter identification in the multi-agent system, thus avoiding the mutual influence between system identification and controller design. Furthermore, because the pseudopartial derivative parameter is time-varying, it can adaptively adjust under different operating environments, making the model applicable to a wide range of environments.

[0105] Step 3: In the face of the situation where the followers do not know the time-varying state information of the leader, based on distributed theory and fixed-time theory, a distributed data-driven fixed-time observer under a non-fully interactive topology network is designed for each follower. This enables effective estimation of the leader's unknown state information and eliminates the communication loop problem among the followers. It provides a guarantee for the design of the fixed-time control method and the formation of a reference tracking trajectory for the global followers, effectively improving the task completion performance of the multi-agent system.

[0106] A distributed leader observer was designed for each follower in a complex interactive network environment. Considering that interactions between agents are not perfect in practical applications, it cannot be guaranteed that each follower agent can effectively obtain the accurate system state of the leader agent in real time, and there may even be situations where followers have no access to the leader's state information at all.

[0107] Therefore, for situations where the leader's state information is unknown, a fixed-time leader observer is designed for each follower as shown below:

[0108]

[0109] in, Indicates the state of the leader The estimated value. The observer parameters satisfy: ,also, All parameters are positive definite and are related to the robustness and fixed-time convergence performance of the observer. It can be obtained that when the parameter matrix satisfies:

[0110]

[0111] in, If it is a positive definite constant, then the follower agent will acquire the state information of the leader agent within a fixed time.

[0112] The following is the proof of the convergence of its observation error.

[0113] Define observation error:

[0114]

[0115] and compact collection and Then we can obtain the following formula:

[0116]

[0117] By utilizing the specific conditions satisfied by the parameter matrix, we can further obtain:

[0118]

[0119] Define the following Lyapunov functions:

[0120]

[0121] The expression for the rate of change can be obtained as follows:

[0122]

[0123] Therefore, we can further conclude that:

[0124]

[0125] for Based on known conditions We can obtain:

[0126]

[0127]

[0128] Therefore, it can be concluded that for ,have:

[0129]

[0130] in,

[0131]

[0132] Furthermore, we have:

[0133]

[0134] in, And we can get:

[0135]

[0136] Therefore, it can be concluded that the global observation error can converge to zero within a fixed time, thus the designed leader observer has fixed-time stability characteristics.

[0137] Step 4: Construct a fixed-time preset performance function to transform the error-constrained tracking problem into an unconstrained stabilization problem through a specific error transformation. Further combine adaptive fixed-time control technology and optimal control technology to construct a data-model hybrid driven fixed-time fault-tolerant controller for heterogeneous multi-agent systems. The proposed controller does not rely on any prior information about the system's mathematical model or communication structure, ensuring that the discrete-time multi-agent system under actuator failure can achieve the reproduction of complex tasks within a fixed time with low energy consumption.

[0138] First, design the following tracking error variables for a multi-agent system:

[0139]

[0140] To achieve performance constraints, a new controlled variable is constructed using a fixed-time pre-set performance method:

[0141]

[0142] in, It is the default performance function, and its expression is:

[0143]

[0144] Among them, parameters All are positive definite constants. Indicates the initial value. This represents the preset fixed convergence time. This is achieved by designing a nonlinear function. New error variables can be obtained. and The equivalent transformation relationship guarantees The control error converges to a finite set in complex environments. Based on this, stabilization... This enables consistent control of preset performance in uncertain multi-agent systems. Furthermore, the fixed-time preset performance function ensures that tracking errors are controlled within a finite, designable timeframe. It can enter the designated region within the system, and the convergence time setting is not constrained by other conditions such as system state, observer gain, and controller parameters.

[0145] By employing the fixed-time theorem and a preset performance control strategy, the control objective of this invention is transformed into: designing a data-driven, fault-tolerant linearized controller using a linearized model and system input / output data, such that the following equation holds:

[0146]

[0147] Typically, due to the existence of unknown uncertainties and faults, a distributed tracking task can be considered achieved when the control error is uniformly bounded. In practical engineering, uniformly bounded convergence is also acceptable. The control objective of this invention can be equivalently stated as: the existence of an ideal input protocol that can complete a uniformly bounded tracking task within a fixed time. Therefore, a pre-defined ideal controller exists, with the following construction:

[0148]

[0149] in, Represents an unknown nonlinear function of the model. This invention represents the convergence error of the actual error with respect to the expected value. It is considered an ideal error value.

[0150] Based on the above analysis, the present invention constructs the following control input:

[0151]

[0152] in, It is an unknown control gain, and:

[0153]

[0154] Based on the above control inputs, the global error can achieve convergence in a fixed time, and the performance indicators meet the constraints of the preset performance function.

[0155] The following design incorporates an adaptive law for the controller parameters. First, we construct a law regarding... Cost function:

[0156]

[0157] in:

[0158]

[0159] And design the model error compensator as shown below:

[0160]

[0161] Based on optimal control theory, its bias is calculated. We can obtain:

[0162]

[0163] Furthermore, we have:

[0164]

[0165] as well as:

[0166]

[0167] Therefore, the following parameters can be obtained. Adaptive regulation law:

[0168]

[0169] And satisfy:

[0170]

[0171] Similarly, design regarding parameters Cost function:

[0172]

[0173] Its adaptive regulation law can be derived as follows:

[0174]

[0175] And satisfy:

[0176]

[0177] The stability of the designed control law is proven below. First, the following expression for the error variable is designed:

[0178]

[0179] in:

[0180]

[0181] Then, define:

[0182]

[0183] We can obtain:

[0184]

[0185] in:

[0186]

[0187] as well as:

[0188]

[0189] Then, we can obtain:

[0190]

[0191] in:

[0192]

[0193] as well as:

[0194]

[0195] Through design parameters This makes the following equation true:

[0196]

[0197] Therefore, we can obtain the consistency tracking error under actuator failure conditions. It is always bounded, which also means that for any k > 0, Both are bounded. Therefore, it can be concluded that the multi-agent system achieves the desired preset performance control within a preset time. The method of this invention only requires simple online adjustment of the adaptive parameters to approximate the ideal controller, effectively reducing algorithm complexity and the performance requirements of industrial control computers, and also facilitating engineering implementation in fault and complex environments.

[0198] Compared to existing data-driven methods, this invention further utilizes model-free optimization techniques, considering the maintainability and reliability of the system, which is of great significance for safety-critical homogeneous and even heterogeneous multi-agent systems. Furthermore, this invention does not strictly rely on system model parameters, and the resulting controller is applicable to various real-time multi-agent systems with external disturbances, model uncertainties, and actuator failures. Simultaneously, the fixed-time preset performance method can constrain the entire evolution of the system's tracking performance, such as overshoot upper limit, overshoot lower limit, convergence speed, and steady-state accuracy. Therefore, by introducing a model error compensator, simplifying the linearized model, and using fixed-time preset performance control, both tracking accuracy and control performance can be effectively guaranteed.

[0199] In this embodiment, one unmanned vehicle acts as the leader, and the followers include two unmanned vehicles (i=1,2) and two unmanned helicopters (i=3,4). The communication topology of each agent in the formation system is as follows: Figure 3 As shown. The leader's initial position is... The initial positions of the other followers are , , ,as well as The controller parameters are set as follows: The settling time is set to To model the actual actuator failure characteristics, the following failure parameters are designed:

[0200]

[0201] To simulate complex external operating conditions, the following external disturbances are designed:

[0202]

[0203] To verify the effectiveness of the heterogeneous multi-agent fault-tolerant formation control method of this invention, simulation verification was performed using the Simulink module in Matlab. Figures 4 to 6 As shown, the controller designed in this invention can effectively adaptively compensate for various actuator faults and external disturbances. Therefore, even if the system model is unknown and the tracking trajectory is a time-varying signal, this heterogeneous multi-agent system can still achieve ideal trajectory tracking results. The tracking accuracy is always maintained within the preset overshoot upper and lower limits, indicating that the designed adaptive specified performance fixed-time controller can effectively compensate for actuator faults and complex uncertainties.

[0204] Depend on Figures 7 to 9It can be seen that the system error always converges within the bounded neighborhood defined by the fixed-time preset performance function, and the error is less than the preset upper bound. This verifies that the control performance of the consistency control system for air-to-ground heterogeneous unmanned systems can be pre-constrained and configured. Simultaneously, it can be deduced that the model error compensator can effectively reconstruct and offset the adverse effects of unknown actuator failures and various unmodeled uncertainties. Therefore, the fixed-time control strategy designed in this invention, by introducing a model error compensator, ensures that the control performance does not depend on the approximation accuracy of the system model or the adaptive estimation accuracy of complex uncertainties. It can directly compensate for complex uncertainties, possessing superior fault tolerance performance and effectively solving the consistency control problem of unknown heterogeneous multi-agent systems under scenarios where external disturbances, model uncertainties, and actuator failures coexist.

[0205] The embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A data-model hybrid driven, fixed-time, fast, fault-tolerant control method for air-to-ground communication, characterized in that, The method is applied to a cross-domain heterogeneous multi-agent system consisting of multiple unmanned helicopters and multiple ground unmanned vehicles, and includes the following steps: (1) Determine that the unmanned vehicle on the road is the leader and the other unmanned vehicles are the followers, and clarify the collaborative configuration and expected tasks of the multi-agent system; use weighted graph theory, symbolic graph and Laplace matrix to quantify the topological changes brought about by information interaction, competitive and cooperative relationship and differences in interests among agents; (2) Analyze the impact of external disturbances and actuator failures on the system and establish a fault description model based on prior knowledge; then, for the distributed error system of the multi-agent system, establish a distributed compact linearization model with unknown parameters by means of the equivalent dynamic linearization method, and use the interaction data between agents to identify and optimize the unknown parameters of the linear model online, and finally form a fault model of the heterogeneous multi-agent system that can describe the main nonlinear complex characteristics. (3) To address the problem that followers cannot obtain the time-varying state information of the leader, based on distributed theory and fixed-time theory, a distributed data-driven fixed-time observer under a non-fully interactive topology network is designed for each follower to achieve effective estimation of the leader's unknown state information, eliminate the communication loop problem existing in each follower, and provide support for the design of subsequent control methods and the formation of a reference tracking trajectory for global followers. (4) Construct a fixed-time preset performance function and transform the tracking problem with error constraints into an unconstrained stabilization problem through a specific error transformation; combine adaptive fixed-time control technology and optimal control technology to construct a data-model hybrid driven fixed-time fault-tolerant controller for heterogeneous multi-agent systems, so as to realize the discrete-time multi-agent system under actuator failure to reproduce complex expected tasks in a fixed time with low energy consumption.

2. The data-model hybrid driven air-to-ground fixed-time fast fault-tolerant control method according to claim 1, characterized in that, The implementation process of step (1) is as follows: Based on graph theory, a directed graph with N followers is used for communication. To describe, among which, and These represent the set of heterogeneous multi-agent systems, the set of communication connection edges, and the connection weight matrix of the communication topology network, respectively. as well as These represent the i-th agent being unable to communicate with the j-th agent and being able to communicate with the i-th agent, respectively. and These represent the i-th follower and the j-th follower being in a friendly cooperative relationship and a competitive relationship, respectively; when At this time, the communication network is a special type of undirected graph network, where the agents communicating with each other have the same weight; the in-degree matrix of the network is represented as... ,in, The Laplace matrix of a communication network is represented as follows: To describe the interaction between leaders and followers, a matrix is ​​designed. To indicate, among which For the connection weight between follower i and leader, when When , it means that the i-th follower can obtain the leader's information, otherwise they cannot.

3. The data-model hybrid driven air-to-ground fixed-time fast fault-tolerant control method according to claim 1, characterized in that, The implementation process of step (2) is as follows: Cross-domain heterogeneous multi-agent systems include an autonomous vehicle leader and There are several followers, including unmanned helicopters and unmanned ground vehicles. To construct a leader-follower consistency controller, the following discrete-time motion model of the leader is constructed: in, and These represent the system state and control input of the leader at time k, respectively. and A system matrix with appropriate dimensions; Secondly, the first The dynamic model of a single follower simplifies to the following discrete-time expression: in, Let K represent the follower's input state and control input data at time k, respectively. Represents an unknown positive definite constant. This represents an unknown function describing the nonlinear characteristics of a follower system; To describe the characteristics of actuator failure, design Using the actual control input and the desired input signal to represent these, we obtain the actuator fault model shown below: in, Denotes an unknown efficiency loss factor, satisfying ; Taking actuator failure into account, the first The dynamic model of a follower, i.e., the fault description model based on prior knowledge, is further expressed as: To facilitate controller design, the following distributed consistency tracking error is constructed: in, Represents a symbolic function. This represents the leader's trajectory, i.e., the desired tracking signal. To fully utilize control input and tracking error data, a dynamic linearization technique is employed to construct the following distributed compact-form linearization model: in, , This represents the time-varying pseudopartial derivative parameter.

4. The fixed-time preset performance fault-tolerant control method for a cross-domain heterogeneous multi-agent system according to claim 1, characterized in that, The distributed data-driven fixed-time observer in the non-fully interactive topology network mentioned in step (3) is: in, This represents the weight of the communication connection between two agents. Indicates the state of the leader The estimated value; the observer parameters satisfy: ,also, All are positive definite parameters to be designed; when the parameter matrix satisfies: in, It is a positive definite constant. yes 3D identity matrix It is a 1-dimensional identity matrix. For a system matrix of appropriate dimension, If we consider the extended Laplace matrix of the communication network, then the follower agent will acquire the state information of the leader agent within a fixed time interval.

5. The data-model hybrid driven air-to-ground fixed-time fast fault-tolerant control method according to claim 1, characterized in that, The implementation process of step (4) is as follows: Design the tracking error variables for the following multi-agent system: To achieve performance constraints, a new controlled variable is constructed using a fixed-time pre-set performance method: in, It is the default performance function, and its expression is: in, All are positive definite constants. Indicates the initial value. This indicates the preset fixed convergence time; Based on the fixed-time theorem and a preset performance control strategy, a data-driven fault-tolerant linearized controller is constructed using a linearized model and system input / output data, such that the following equation holds: To design a controller with pre-defined ideal performance for completing a consistent bounded tracking task within a fixed time, its construction is as follows: in, Represents an unknown nonlinear function of the model. This indicates the convergence error of the actual error with respect to the expected value. Assuming an ideal error value, construct the following control input: in, It is an unknown control gain, and: Based on the above control inputs, the global error achieves fixed-time convergence, and the performance indicators meet the constraints of the preset performance function. The following design incorporates an adaptive law for the controller parameters: First, construct a law regarding... Cost function: in: And design the model error compensator as shown below: Based on optimal control theory, its bias is calculated. ,get: Furthermore, we have: as well as: The following parameters were obtained. Adaptive regulation law: And satisfy: Design about parameters Cost function: Its adaptive regulation law is derived as follows: And satisfy: 。

Images