High-efficiency intelligent networked inspection robot dog cluster cooperative control method

Abstract

The application discloses a high-efficiency intelligent networked inspection robot dog cluster cooperative control method, and relates to the technical field of intelligent networked inspection robot dog cluster cooperative control; the method comprises the following steps: constructing a networked inspection robot dog cluster system dynamics model composed of a plurality of four-legged robot dog single rigid body dynamics models affected by uncertain disturbances; based on the networked inspection robot dog cluster system dynamics model, an equivalent motion tracking error system model is established; according to the equivalent motion tracking error system model, an event-triggered distributed model predictive control method with adaptive adjustment of a prediction time domain is designed; and the recursive feasibility of the control method and the stability of the closed-loop system are verified. The method in the application realizes safe and stable control of the intelligent networked inspection robot dog cluster, not only maintains the cooperative consistency of the robot dog cluster, but also significantly reduces the communication and calculation load of the overall network, and significantly improves the work efficiency, system robustness and scene adaptability in complex inspection tasks.

Description

Technical Field

[0001] This invention belongs to the field of intelligent connected inspection robot dog cluster collaborative control technology, and particularly relates to an efficient intelligent connected inspection robot dog cluster collaborative control method. Background Technology

[0002] In numerous fields such as industrial facilities, urban security, and disaster relief, the demand for routine and refined inspections in complex environments is increasingly urgent. However, traditional inspection methods based on fixed monitoring, human intervention, and single robotic dogs face many bottlenecks. When facing unstructured terrain, traditional wheeled or tracked robotic dogs struggle to navigate, while quadrupedal robotic dogs offer greater flexibility. However, even in vast, multi-dimensional inspection spaces (such as large factories, substations, or disaster sites), a single quadrupedal robotic dog has limitations. Its perception range, continuous operating time, computing power, and load are all restricted; furthermore, a single quadrupedal robotic dog system has poor robustness, and a failure will interrupt the entire inspection task. Therefore, many scholars have proposed forming a cluster of multiple quadrupedal robotic dogs through networked collaborative control, making them an intelligent and efficient organic whole. Research on the collaborative control of intelligent networked inspection robotic dog clusters represents a leap from quantitative to qualitative change in system capabilities. It is not only a natural result of technological progress but also a key enabling technology for unlocking the next generation of intelligent inspection systems and reshaping future industrial and social operating models, possessing profound research value and practical significance.

[0003] Currently, the collaborative control algorithms for intelligent connected inspection robot dog swarms are developing rapidly, achieving numerous results. Model predictive control-based algorithms, through distributed frameworks and consensus protocols, enable robot dog swarms to achieve obstacle avoidance, navigation, and gait synchronization. Bionic collaborative control algorithms draw on the "motor coordination" model of animal movement, providing a stable individual foundation for swarm collaboration. Distributed collaborative control algorithms improve the flexibility and robustness of the swarm. Reinforcement learning-based algorithms allow robot dogs to autonomously learn optimal collaborative strategies. However, few studies have considered the waste of communication and computing resources caused by traditional time-triggered control in large-scale, high-efficiency, and long-endurance intelligent inspection robot dog swarm systems. Event-triggered mechanisms have significant advantages in saving communication resources, reducing computational load, extending system endurance, and enhancing system scalability. In the future, quadruped robot dog swarm collaborative motion control algorithms will develop towards intelligence, autonomy, and integration to adapt to complex and changing environments and diverse task requirements.

[0004] Traditional control algorithms employ event-triggered mechanisms to control agents, which are typically based on simplified point mass or integrator models, resulting in low dependence on dynamic characteristics for their control tasks. Quadruped robot dog swarms are complex swarm systems with highly nonlinear, strongly coupled dynamics and stringent physical constraints. This invention addresses these complex swarm systems and, to solve the unique challenges in specific application scenarios, adjusts and redesigns the theory.

[0005] Therefore, this invention proposes an event-triggered distributed model predictive control method based on an adaptive adjustment prediction time domain of an intelligent inspection robot dog cluster system. This method effectively reduces the impact of external interference on the movement of the inspection robot dogs, achieves safe and stable control of the intelligent connected inspection robot dog cluster, maintains the collaborative consistency of the robot dog cluster, significantly reduces the communication and computing load of the overall network, and significantly improves the work efficiency, system robustness, and scenario adaptability in complex inspection tasks. Summary of the Invention

[0006] The purpose of this invention is to provide an efficient intelligent connected inspection robot dog cluster collaborative control method to solve the problems of communication and computing resource waste caused by traditional time-triggered control in large-scale, high-efficiency, and long-endurance intelligent inspection robot dog cluster systems, as mentioned in the background art.

[0007] To achieve the above objectives, the present invention employs the following technical solution:

[0008] This invention proposes a highly efficient intelligent connected inspection robot dog cluster collaborative control method, comprising the following steps:

[0009] S1. Construct a dynamic model of the networked inspection robot dog cluster system. The dynamic model of the networked inspection robot dog cluster system consists of several single rigid body dynamic models of quadruped robot dogs affected by uncertain disturbances.

[0010] S2. Based on the dynamic model of the networked inspection robot dog cluster system, establish an equivalent motion tracking error system model and a corresponding nominal tracking error system model;

[0011] S3. Based on the equivalent motion tracking error system model, design the optimization control problem of the quadruped robot dog control system based on the prediction time domain and trigger time, and solve the optimization control problem as the model predictive controller; design an event-triggered distributed model predictive control method that adaptively adjusts the prediction time domain based on the prediction time domain adjustment strategy and event triggering mechanism.

[0012] S4. Verify the recursive feasibility and closed-loop system stability of the event-triggered distributed model predictive control method for adaptive adjustment of the prediction time domain.

[0013] Preferably, the single rigid body dynamic model of the quadruped robot dog in S1 is as follows:

[0014] The disturbed dynamic equations of the quadruped robot dog are as follows:

[0015]

[0016]

[0017]

[0018]

[0019] in, Indicates the first The acceleration of the center of mass of a quadruped robot dog Indicates the first The mass of the quadruped robot dog, g represents the acceleration due to gravity. Indicates the first The first four-legged robotic dog The reaction force at the foot of one leg; For the first The angular velocity of the center of mass of a quadruped robot dog in the world coordinate system. Indicates the first The first four-legged robotic dog The distance vector of a leg from the point of ground reaction force to its center of mass; This is the roll angle. The pitch angle, Yaw angle Indicates the first A four-legged robotic dog circled around Z Rotation of axis in the positive direction angle, Indicates the first The rotation matrix from the body coordinate system of the quadruped robot dog to the world coordinate system; For the first The inertial tensor in the world coordinate system when the quadruped robot dog moves. For the first The inertial tensor in the body coordinate system of a quadruped robot dog; , For the first The disturbance term of external environment and internal system uncertainty experienced by a quadruped robot dog;

[0020] The first Euler angles of a four-legged robotic dog Centroid coordinates ,speed and angular velocity As the corresponding quadruped robot dog system state, and writing the gravitational acceleration into the coefficient matrix, we obtain the first... A quadruped robot dog system state variables The ground reaction force is defined as the control input, i.e. , obtained the State-space equations for a quadruped robot dog system:

[0021]

[0022]

[0023] in, and The first The status and control inputs of a quadruped robot dog system. It is a disturbance term, and satisfies ,in, Given constants; matrix .

[0024] Furthermore, the constraint conditions for the friction cones at the foot ends of the quadruped robot dog are as follows:

[0025]

[0026] Minimum force Set to 0 and the foot force is set within the range of the friction cone, that is:

[0027]

[0028]

[0029] in, Indicates the first The first four-legged robotic dog The reaction force at the foot of each leg.

[0030] Preferably, the equivalent motion tracking error system model in S2 is as follows:

[0031] Assume the first The reference trajectory of a quadruped robot dog under strong interference is one where the robot body is highly stable and horizontally non-rotating, i.e., the reference state is... The tracking error is defined as:

[0032]

[0033] Therefore, the tracking error system model and equivalent constraints are obtained:

[0034]

[0035]

[0036] The expression for the nominal tracking error system model is:

[0037]

[0038] And there exists a matrix , making the matrix It's by Hurwitz.

[0039] Preferably, the event-triggered distributed model predictive control method for adaptively adjusting the prediction time domain in S3 is as follows:

[0040] S301. Design an optimization control problem based on the input trajectory and predicted state trajectory of the nominal tracking error system model. By solving the optimization control problem, the control input of the quadruped robot dog control system is obtained, and the model predictive controller is obtained.

[0041] S302. The prediction time domain adjustment strategy adopts an adaptive adjustment prediction time domain mechanism, and designs a gradually shortened prediction time domain to make the quadruped robot dog control system approach the terminal domain to maintain the feasibility of the system.

[0042] S303. Design event triggering conditions based on state prediction deviation.

[0043] Preferably, the optimization control problem in S301 is designed as follows:

[0044] The first The quadruped robot dog control system serves as a subsystem of the networked inspection robot dog cluster system. Define the prediction time domain and trigger time series are ,in Subsystem At the trigger time The local optimization problem is expressed as:

[0045]

[0046]

[0047]

[0048]

[0049]

[0050] in, and Representing the nominal subsystems Feasible input trajectories and corresponding predicted state trajectories, superscript Optimal solution; Representation Subsystem Neighbor subsystem exist The predicted state at any given moment; Representation Subsystem Terminal constraints;

[0051] Coupling cost function Represented as:

[0052]

[0053] in, The weight matrix is ​​used as the input for both state and control. For consistency weight matrix, This is the terminal weight matrix;

[0054] Subsystem The control input is obtained by solving an optimization problem. The design is as follows:

[0055]

[0056] in, Indicates the first A model of a quadruped robot dog predicts control inputs.

[0057] A quadruped robot dog cluster consists of multiple independent quadruped robot dogs, each of which is considered a subsystem. i Each quadruped robot dog is modeled in the same way, and each quadruped robot dog subsystem... i Each has an independent controller. i Each controller i The control methods employed are all adaptive, time-domain-triggered, distributed model predictive control methods, applicable to each quadruped robot dog in the cluster. This control method considers the motion consistency of each robot dog and, when solving the local optimization problem, designs a cost function... This is to ensure the synchronization of the posture of each quadruped robot dog.

[0058] Preferably, the adaptive adjustment prediction time-domain mechanism in S302 is as follows:

[0059] First of all, in the current situation At the trigger point, the shortest prediction time range that guarantees the feasibility of the optimal control problem is:

[0060]

[0061] Subsequently, the time-domain scale of the shrinking prediction Designed as follows:

[0062]

[0063] in, It is an adjustment parameter to ensure feasibility; The trigger parameter is defined and the condition is met. Therefore, we obtain The predicted time domain will show a decreasing or non-increasing trend;

[0064] In the networked inspection robot dog cluster system, all four-legged robot dogs have the same initial prediction time domain setting, which is... .

[0065] Preferably, the event triggering mechanism is as follows:

[0066] The event triggering conditions are designed based on state prediction deviations, as follows:

[0067]

[0068] in, As the trigger threshold, Define the trigger parameters to be designed; ; These are the tuning parameters;

[0069] The next trigger time is:

[0070]

[0071] in, Therefore, the subsystem When the state has not entered the terminal domain, the lower and upper bounds of the execution time of the model predictive controller are:

[0072]

[0073] in, The model predicts the execution time of the controller.

[0074] Preferably, the verification of recursion feasibility in step S4 is as follows:

[0075] Known and And simultaneously satisfy and ,get :

[0076] For closed-loop networked inspection robot dog cluster system If the following conditions are met: i) ;ii) ;iii) ;

[0077] This ensures the recursive feasibility of the proposed control method.

[0078] Consider the following control inputs:

[0079]

[0080] In time interval In China, according to ,get:

[0081]

[0082] Applying the Gronwall-Bellman inequality, we obtain:

[0083]

[0084] Will Substituting into the above equation, and combining with condition i) and the triangle inequality, we get:

[0085]

[0086] In addition, within the time interval Within, according to the comparison principle, condition ii), and the above equation, we obtain:

[0087]

[0088] This yields the state. Meet terminal constraints;

[0089] Next, regarding the time interval Obviously satisfied For time intervals It can be deduced that This yields the candidate control inputs. Satisfy control input constraints ;

[0090] This allows us to obtain the control input. exist The triggering time is a feasible solution, and the feasibility of the recursion has been verified.

[0091] Preferably, the stability verification of the closed-loop system in S4 is as follows:

[0092] Under the condition that recursion feasibility is satisfied, the inspection robot dog cluster system under adaptive time-domain event-triggered distributed model predictive control will remain stable if the following conditions are met:

[0093]

[0094]

[0095]

[0096] If the conditions are met ,in ,as well as Then, the closed-loop system will remain in the terminal domain after entering the terminal domain.

[0097] Furthermore, the system state will converge to the set It is given by the following formula:

[0098]

[0099] in,

[0100]

[0101] First, consider Let the optimal cost function be denoted as You can get satisfy:

[0102]

[0103] according to We can obtain:

[0104]

[0105] Based on the above analysis, we can conclude that:

[0106]

[0107] in,

[0108]

[0109]

[0110]

[0111]

[0112] for Using Holder's inequality, we can obtain:

[0113]

[0114] exist and In, by using and It can be observed that:

[0115]

[0116] For the second component By using the formula and We can obtain:

[0117]

[0118] For the third component According to the triangle inequality, we know that:

[0119]

[0120] For the fourth component We can obtain:

[0121]

[0122] according to and We can obtain:

[0123]

[0124] according to We can obtain:

[0125]

[0126] Next, regarding the situation We can obtain:

[0127]

[0128] Regarding the situation We can obtain:

[0129]

[0130] The above two situations can be summarized as follows:

[0131]

[0132] Based on the above, we can conclude that:

[0133]

[0134] Four components Adding them together yields the result; therefore, it can be concluded that the state in the system will enter the terminal domain within a finite time. ;

[0135] Consider when When, choose For Lyapunov functions, we can obtain:

[0136]

[0137] Will and Substituting into the above equation, we get:

[0138]

[0139] Next, based on the design process of the dual-mode event triggering mechanism, we can obtain:

[0140]

[0141] Based on the triggering conditions:

[0142]

[0143] Therefore, we can obtain It is an invariant domain, meaning that once the system state enters the terminal domain, it remains within the terminal domain. In summary, the stability verification of the closed-loop system is complete.

[0144] Compared with the prior art, the beneficial effects of the present invention are:

[0145] (1) The efficient intelligent networked inspection robot dog cluster collaborative control method in this invention, by introducing a predictive time-domain adaptive adjustment mechanism and an event-triggered communication mechanism, realizes the networked safe and stable control of the intelligent inspection robot dog cluster. It not only maintains the collaborative consistency of the four-legged robot dog cluster, but also significantly reduces the communication and computing load of the overall network, and significantly improves the work efficiency, system robustness and scenario adaptability in complex inspection tasks.

[0146] (2) The dynamic model of the robot dog cluster system designed in this invention conforms to the characteristics of a quadruped robot dog cluster, and has highly nonlinear complex dynamics (each robot dog is an underactuated, strongly coupled multibody system that satisfies the contact dynamics between the foot and the ground), strict physical constraints (foot friction cone constraints, etc.), and terrain adaptation requirements (the cluster needs to move collaboratively in an unstructured environment to meet the requirements of control accuracy and real-time performance).

[0147] (3) The adaptive adjustment mechanism of the prediction time domain in this invention is different from the adaptive adjustment of the prediction time domain in traditional control algorithms, which is mostly based on the error of the system state or the complexity of future dynamics. This invention deeply couples it with the characteristics of the robot dog. Instead of setting a unified prediction time domain for the entire cluster, it dynamically adjusts the prediction time domain for each robot dog according to its current state.

[0148] (4) The event-triggered communication mechanism in this invention differs from the traditional control algorithm where the event triggering condition is usually based solely on the norm error of the agent's state. In event-triggered communication, this invention designs a composite triggering condition that comprehensively considers both the "kinematic consistency requirements" and the "communication resource bottleneck" of the robot dog cluster. Attached Figure Description

[0149] Figure 1 This is a flowchart of the efficient intelligent connected inspection robot dog cluster collaborative control method in this invention;

[0150] Figure 2 This is a schematic diagram of the adaptive adjustment and prediction time-domain event-triggered distributed model predictive control principle of the intelligent inspection robot dog cluster system in this invention.

[0151] Figure 3 This is a schematic diagram of the Euler angles and position trajectory of the intelligent inspection robot dog under the control method of the present invention;

[0152] Figure 4 This is a schematic diagram of the angular velocity and linear velocity trajectory of the intelligent inspection robot dog under the control method of the present invention;

[0153] Figure 5 This is a schematic diagram of the force trajectory at the front paws of the intelligent inspection robot dog under the control method of the present invention;

[0154] Figure 6 This is a schematic diagram of the force trajectory at the rear foot of the intelligent inspection robot dog under the control method of the present invention;

[0155] Figure 7 This is a schematic diagram of the controller triggering time for the efficient intelligent connected inspection robot dog cluster collaborative control method in this invention;

[0156] Figure 8 This is a schematic diagram of the predicted time-domain evolution process of the efficient intelligent connected inspection robot dog cluster collaborative control method in this invention. Detailed Implementation

[0157] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0158] Example 1:

[0159] See Figure 1A highly efficient and intelligent connected inspection robot dog cluster collaborative control method, specifically an adaptive adjustment and prediction time-domain event-triggered distributed model predictive control method for an intelligent connected inspection robot dog cluster system, includes the following steps:

[0160] S1. Construct a cluster dynamics model consisting of five networked inspection robot dogs affected by uncertain disturbances.

[0161] Specifically, the control method is based on a single rigid body dynamics model. This model simplifies the quadruped robot dog into a rigid body moving in space, treating the interaction force between its feet and the ground as an external force. This is achieved by defining a world coordinate system. and body coordinate system Furthermore, by applying the Newton-Euler equations, the dynamic relationship between the robot dog's center of mass acceleration and the force applied to its feet was directly established:

[0162] (1)

[0163] in, Indicates the first The position of the center of mass of a quadruped robot dog in the world coordinate system. Indicates the first The weight of a quadruped robot dog Represents gravitational acceleration. Indicates the first The first four-legged robotic dog The reaction force at the foot of each leg.

[0164] set up Let the inertia tensor in the robot dog's body coordinate system be the mass of the quadruped robot dog. ,length ,width and height Conclusion:

[0165] (2)

[0166] When the quadruped robot dog moves, the inertia tensor in the world coordinate system is:

[0167] (3)

[0168] in, Indicates the first The rotation matrix from the body coordinate system of the quadruped robot dog to the world coordinate system.

[0169] Based on the angular momentum theorem, the vector expression for the moment of inertia of the quadruped robot dog is obtained:

[0170] (4)

[0171] (5)

[0172] in, For the first The angular velocity of the center of mass of a quadruped robot dog in the world coordinate system. Indicates the first The first four-legged robotic dog The distance vector of each leg from the point of ground reaction to the center of mass is the position of the center of mass obtained by the state estimator minus the distance of the first leg. The position of one leg For any vector , Indicates satisfaction The antisymmetric matrix;

[0173] The robot dog's pose is typically achieved using ZYX Euler angles in the body coordinate system. It means that among them This is the roll angle. The pitch angle, Let yaw be the angle of rotation; from this, the expression for the rotation matrix from the body coordinate system to the world coordinate system is:

[0174] (6)

[0175] in, , , , indicating around Rotation of axis in the positive direction angle.

[0176] in,

[0177] (7)

[0178] Next, using the rotation matrix, the angular velocity of the robot dog's center of mass in the world coordinate system is obtained based on the robot dog's attitude Euler angles:

[0179] (8)

[0180] in:

[0181] (9)

[0182] Considering only the case where the quadruped robot dog walks on the ground, and assuming the robot dog's pitch angle is not 90° (i.e. ), for the mapping matrix Inverse mapping yields the reverse mapping relationship:

[0183] (10)

[0184] To ensure stability during movement, the quadruped robot dog has a pitch angle. and roll angle Approaching zero; simultaneously, the angle between the Z-axis of the body coordinate system and the Z-axis of the world coordinate system is almost zero; simplifying the above expression yields:

[0185] (11)

[0186] According to Euler's formula, considering that the angular velocity of the robot dog during its movement is relatively small, it can be ignored. From the item, we can obtain:

[0187] (12)

[0188] Considering the multi-source disturbances caused by external environmental and internal system uncertainties affecting the robot dog, and given that the robot dog's pitch and roll angles are close to zero, the disturbed dynamic equations for the quadruped robot dog can be obtained as follows:

[0189] (13)

[0190] (14)

[0191] (15)

[0192] (16)

[0193] The Euler angles, center-of-mass coordinates, velocity, and angular velocity of the quadruped robot dog are taken as the system state, and the gravitational acceleration is written into the coefficient matrix to obtain the system state variables. The ground reaction force is defined as the control input, i.e. The state-space equations of the system can be obtained as follows:

[0194] (17)

[0195] (18)

[0196] in, and These are the system's state and control inputs, respectively. It is a disturbance term, and satisfies ,in, Given constants; matrix Furthermore, the constraint conditions for the friction cones at the foot ends of the quadruped robot dog are as follows:

[0197] (19)

[0198] Since the force exerted by the ground on the feet of the quadruped robot dog in the vertical direction cannot be negative, the minimum force is... Set to 0; furthermore, to prevent the robot dog's feet from slipping while walking, the force applied to the feet must be within the range of the friction cone, i.e.:

[0199] (20)

[0200] Therefore, we can conclude that:

[0201] (twenty one)

[0202] (twenty two)

[0203] S2. Based on the dynamic model of the networked inspection robot dog cluster, establish an equivalent motion tracking error system model.

[0204] Assuming the quadruped robot dog's reference trajectory under strong interference is one where the robot is highly stable and horizontally non-rotating, i.e., the reference state is... The tracking error is defined as:

[0205] (twenty three)

[0206] Therefore, the tracking error system and equivalent constraints are obtained:

[0207] (twenty four)

[0208] (25)

[0209] The expression for the nominal tracking error system is:

[0210] (26)

[0211] And there exists a matrix , making the matrix It's by Hurwitz.

[0212] S3. Based on the equivalent motion tracking error system model of the networked inspection robot dog cluster, design an adaptive adjustment prediction time-domain event-triggered distributed model predictive control method, such as... Figure 2 As shown.

[0213] S3.1 Description of the optimization control problem:

[0214] Define the prediction time domain and trigger time series are ,in The system at the trigger time The local optimization problem can be expressed as:

[0215] (27)

[0216] (28)

[0217] (29)

[0218] (30)

[0219] (31)

[0220] in, and Representing the nominal subsystems Feasible input trajectories and corresponding predicted state trajectories, superscript Optimal solution; Representation Subsystem Neighbor subsystem exist The predicted state at any given moment; Represents system terminal constraints; coupling cost function Represented as:

[0221] (32)

[0222] in, The weight matrix is ​​used as the input for both state and control. For consistency weight matrix, This is the terminal weight matrix;

[0223] Subsystem The control input is obtained by solving an optimization problem. The design is as follows:

[0224] (33)

[0225] S3.2 Adaptive Adjustment Prediction Time Domain Mechanism:

[0226] Typically, as the system approaches the terminal domain, a shorter prediction time can maintain system feasibility; therefore, a gradually shortening prediction time is designed. Firstly, in the current... At the trigger point, the shortest prediction time range that guarantees the feasibility of the optimal control problem is:

[0227] (34)

[0228] Subsequently, the time-domain scale of the shrinking prediction Designed as:

[0229] (35)

[0230] in, It is an adjustment parameter to ensure feasibility; The trigger parameter is defined and the condition is met. Therefore, according to the above formula, we can obtain... The predicted time domain will show a decreasing or non-increasing trend, such as To simplify calculations, all quadruped robot dogs have the same initial prediction time domain setting, which is... .

[0231] S3.3 Event Triggering Mechanism:

[0232] Design event triggering conditions based on state prediction deviation:

[0233] (36)

[0234] in, As the trigger threshold, Define the trigger parameters to be designed; ; These are the tuning parameters;

[0235] The next trigger time is:

[0236] (37)

[0237] in, Therefore, when the system state has not entered the terminal domain, the lower and upper bounds of the controller's execution time are: .

[0238] S4. Verify the recursive feasibility of the control method and the stability of the closed-loop system, specifically including the following steps:

[0239] S4.1 Recursive Feasibility Analysis:

[0240] Known and And simultaneously satisfy and You can get :

[0241] For closed-loop inspection robot dog cluster system The recursive feasibility of the proposed control strategy can be guaranteed if the following conditions are met: i) ;ii) ;iii) ;

[0242] Consider the following control inputs:

[0243] (38)

[0244] In the time interval In China, according to ,get:

[0245] (39)

[0246] Applying the Gronwall-Bellman inequality, we obtain:

[0247] (40)

[0248] Will Substituting into the above equation, and combining with condition i) and the triangle inequality, we get:

[0249] (41)

[0250] In addition, within the time interval Within, according to the comparison principle, condition ii), and the above equation, we obtain:

[0251] (42)

[0252] This yields the state. Meet terminal constraints;

[0253] Next, regarding the time interval Obviously satisfied For time intervals It can be deduced that This yields the candidate control inputs. Satisfy control input constraints ;

[0254] This allows us to obtain the control input. exist The triggering time is a feasible solution, and the feasibility of the recursion has been verified.

[0255] S4.2, Stability analysis of closed-loop system:

[0256] Under the condition that recursion feasibility is satisfied, the inspection robot dog cluster system under adaptive time-domain event-triggered distributed model predictive control will remain stable if the following conditions are met:

[0257] (43)

[0258] in,

[0259] (44)

[0260] (45)

[0261] In addition, if the conditions are met ,in ,as well as Therefore, it can be concluded that the closed-loop system will remain within the terminal domain after entering the terminal domain;

[0262] Furthermore, the system state will converge to the set It is given by the following formula:

[0263] (46)

[0264] in,

[0265] (47)

[0266] First, consider Let the optimal cost function be denoted as You can get satisfy:

[0267] (48)

[0268] according to We can obtain:

[0269] (49)

[0270] Based on the above analysis, we can conclude that:

[0271] (50)

[0272] in,

[0273] (51)

[0274] (52)

[0275] (53)

[0276] (54)

[0277] for Using Holder's inequality, we can obtain:

[0278] (55)

[0279] exist and In, by using and It can be observed that:

[0280] (56)

[0281] For the second component By using the formula and We can obtain:

[0282] (57)

[0283] For the third component According to the triangle inequality, we know that:

[0284] (58)

[0285] For the fourth component We can obtain:

[0286] (59)

[0287] according to and We can obtain:

[0288] (60)

[0289] according to We can obtain:

[0290] (61)

[0291] Next, regarding the situation We can obtain:

[0292] (62)

[0293] Regarding the situation We can obtain:

[0294] (63)

[0295] The above two situations can be summarized as follows:

[0296] (64)

[0297] Based on the above, we can conclude that:

[0298] (65)

[0299] Four components Adding them together yields the result; therefore, it can be concluded that the state in the system will enter the terminal domain within a finite time. ;

[0300] Consider when When, choose For Lyapunov functions, we can obtain:

[0301] (66)

[0302] Will and Substituting into the above equation, we get:

[0303] (67)

[0304] Next, based on the design process of the dual-mode event triggering mechanism, we can obtain:

[0305] (68)

[0306] Based on the triggering conditions:

[0307] (69)

[0308] Therefore, we can obtain It is an invariant domain, meaning that once the system state enters the terminal domain, it remains within the terminal domain. In summary, the stability verification of the closed-loop system is complete.

[0309] Experimental verification:

[0310] To verify the efficient intelligent connected inspection robot dog cluster system collaborative control method provided in this embodiment, MATLAB was used for simulation experiments, and detailed explanations are provided below:

[0311] The quadrupedal networked inspection robot dog cluster model provided in this embodiment is designed with an adaptive adjustment prediction time domain event-triggered distributed model predictive control strategy. By introducing a prediction time domain adaptive adjustment mechanism and an event-triggered communication mechanism, the networked safe and stable control of the inspection robot dog cluster is realized. This not only maintains the collaborative consistency of the robot dog cluster, but also significantly reduces the communication and computing load of the overall network, and significantly improves the operation efficiency, system robustness and scenario adaptability in complex inspection tasks.

[0312] The quadruped robot dog swarm system consists of five identical quadruped robot dogs. Its parameters are set as follows: in the simulation, the robot dog's body is simplified to a rectangle, the leg mass is ignored, the robot dog's center of mass is at the geometric center of gravity, and its mass is [value missing]. Its length, width and height are respectively , ,as well as The acceleration due to gravity is The coefficient of friction is Set the minimum force at the foot. Maximum force at the foot Inertial tensor in body coordinate system Assume that the five intelligent inspection robot dogs have different initial states; initially, the four-legged robot dogs are at rest, and the force exerted vertically upward on each foot is... In addition, apply perturbation acceleration and disturbance angular acceleration The control objective is to maintain the stable posture of the quadruped robot dog affected by multiple disturbances and to ensure the consistency of robot dog cluster control.

[0313] The parameter design of the efficient networked collaborative control method for intelligent inspection robot dogs is as follows: State weight matrix Consistency weight matrix Input weight matrix It is a diagonal matrix with elements of 0.05; the sampling time is... Total simulation duration , , , , , , .

[0314] Based on the above parameters, the control method proposed in this invention was simulated and verified. The corresponding simulation results are as follows: Figures 3-6 As shown. Among them, Figure 3 and Figure 4 The state trajectories of five intelligent inspection robot dog clusters under the proposed adaptive adjustment prediction time-domain event-triggered distributed model predictive control strategy are shown. Figure 5 and Figure 6 The force trajectories of the four feet of the five intelligent inspection robot dog clusters in the XYZ axis directions under the proposed control method are shown, and are constrained within the anti-slip force constraint. Figure 7 This indicates the controller trigger time of the proposed control strategy; Figure 8 The predicted time-domain evolution process of the proposed control strategy is demonstrated.

[0315] The above analysis verifies the effectiveness of the adaptive adjustment and prediction time-domain event-triggered distributed model predictive control method of the intelligent inspection robot dog cluster system provided in this embodiment. It realizes the networked, safe, and stable control of the intelligent inspection robot dog cluster, not only maintaining the collaborative consistency of the four-legged robot dog cluster, but also significantly reducing the communication and computing load of the overall network, and significantly improving the work efficiency, system robustness, and scenario adaptability in complex inspection tasks.

[0316] The above description is only for the purpose of helping to understand the method and core essence of the present invention, but the scope of protection of the present invention is not limited thereto. For those skilled in the art, any equivalent substitutions or modifications made within the technical scope disclosed in the present invention, based on the technical solution and inventive concept, should be covered within the scope of protection of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A highly efficient and intelligent networked inspection robot dog cluster collaborative control method, characterized in that, Includes the following steps: S1. Construct a dynamic model of the networked inspection robot dog cluster system. The dynamic model of the networked inspection robot dog cluster system consists of several single rigid body dynamic models of quadruped robot dogs affected by uncertain disturbances. S2. Based on the dynamic model of the networked inspection robot dog cluster system, establish an equivalent motion tracking error system model and a corresponding nominal tracking error system model; S3. Based on the equivalent motion tracking error system model, design the optimization control problem of the quadruped robot dog control system based on the prediction time domain and trigger time, and solve the optimization control problem as the model predictive controller; design an event-triggered distributed model predictive control method that adaptively adjusts the prediction time domain based on the prediction time domain adjustment strategy and event triggering mechanism. An event-triggered distributed model predictive control method for adaptive adjustment of the prediction time domain is described below: S301. Design an optimization control problem based on the input trajectory and predicted state trajectory of the nominal tracking error system model. By solving the optimization control problem, the control input of the quadruped robot dog control system is obtained, and the model predictive controller is obtained. S302. The prediction time domain adjustment strategy adopts an adaptive adjustment prediction time domain mechanism, and designs a gradually shortened prediction time domain to make the quadruped robot dog control system approach the terminal domain to maintain the feasibility of the system. The adaptive adjustment mechanism for the prediction time domain is as follows: First of all, in the current situation At the trigger point, the shortest prediction time range that guarantees the feasibility of the optimal control problem is: in, Indicates the nominal subsystem The predicted state trajectory, superscript Optimal solution; Representation Subsystem Terminal constraints; Subsequently, the time-domain scale of the prediction was reduced. Designed as follows: in, It is an adjustment parameter to ensure feasibility; The trigger parameter is defined and the condition is met. Therefore, we obtain The predicted time domain will show a decreasing or non-increasing trend; In the networked inspection robot dog cluster system, all four-legged robot dogs have the same initial prediction time domain setting, which is... ; S303. Design event triggering conditions based on state prediction deviation; S4. Verify the recursive feasibility and closed-loop system stability of the event-triggered distributed model predictive control method for adaptive adjustment of the prediction time domain.

2. The efficient intelligent connected inspection robot dog cluster collaborative control method according to claim 1, characterized in that, The single rigid body dynamics model of the quadruped robot dog in S1 is as follows: The disturbed dynamic equations of the quadruped robot dog are as follows: in, Indicates the first The acceleration of the center of mass of a quadruped robot dog Indicates the first The mass of the quadruped robot dog, g represents the acceleration due to gravity. Indicates the first The first four-legged robotic dog The reaction force at the foot of one leg; For the first The angular velocity of the center of mass of a quadruped robot dog in the world coordinate system. Indicates the first The first four-legged robotic dog The distance vector of a leg from the point of ground reaction force to its center of mass; This is the roll angle. The pitch angle, Yaw angle Indicates the first A four-legged robotic dog circled around Z Rotation of axis in the positive direction angle, Indicates the first The rotation matrix from the body coordinate system of the quadruped robot dog to the world coordinate system; For the first The inertial tensor in the world coordinate system when the quadruped robot dog moves. For the first The inertial tensor in the body coordinate system of a quadruped robot dog; , For the first The disturbance term of external environment and internal system uncertainty experienced by a quadruped robot dog; The first Euler angles of a four-legged robotic dog Centroid coordinates ,speed and angular velocity As the corresponding quadruped robot dog system state, and writing the gravitational acceleration into the coefficient matrix, we obtain the first... A quadruped robot dog system state variables The ground reaction force is defined as the control input, i.e. , obtained the State-space equations for a quadruped robot dog system: in, and The first The status and control inputs of a quadruped robot dog system. It is a disturbance term, and satisfies ,in, Given constants; matrix .

3. The efficient intelligent networked inspection robot dog cluster collaborative control method according to claim 2, characterized in that, The constraint conditions for the foot friction cones of the quadruped robot dog are as follows: Minimum force Set to 0 and the foot force is set within the range of the friction cone, that is: in, Indicates the first The first four-legged robotic dog The reaction force at the foot of each leg.

4. The efficient intelligent connected inspection robot dog cluster collaborative control method according to claim 2 or 3, characterized in that, The equivalent motion tracking error system model in S2 is as follows: Assume the first The reference trajectory of a quadruped robot dog under strong interference is one where the robot body is highly stable and horizontally non-rotating, i.e., the reference state is... The tracking error is defined as: Therefore, the tracking error system model and equivalent constraints are obtained: The expression for the nominal tracking error system model is: And there exists a matrix , making the matrix It's by Hurwitz.

5. The efficient intelligent networked inspection robot dog cluster collaborative control method according to claim 4, characterized in that, The design optimization control problem in S301 is as follows: The first The quadruped robot dog control system serves as a subsystem of the networked inspection robot dog cluster system. Define the prediction time domain and trigger time series are ,in Subsystem At the trigger time The local optimization problem is expressed as: in, and Representing the nominal subsystems Feasible input trajectories and corresponding predicted state trajectories, superscript Optimal solution; Representation Subsystem Neighbor subsystem exist The predicted state at any given moment; Representation Subsystem Terminal constraints; Coupling cost function Represented as: in, The weight matrix is ​​used as the input for both state and control. For consistency weight matrix, This is the terminal weight matrix; Subsystem The control input is obtained by solving an optimization problem. The design is as follows: in, Indicates the first A model of a quadruped robot dog predicts control inputs.

6. The efficient intelligent networked inspection robot dog cluster collaborative control method according to claim 5, characterized in that, The event triggering mechanism is as follows: The event triggering conditions are designed based on state prediction deviations, as follows: in, As the trigger threshold, Define the trigger parameters to be designed; ; These are the tuning parameters; The next trigger time is: in, Therefore, the subsystem When the state has not entered the terminal domain, the lower and upper bounds of the execution time of the model predictive controller are: in, The model predicts the execution time of the controller.

7. The efficient intelligent networked inspection robot dog cluster collaborative control method according to claim 6, characterized in that, The feasibility of recursion is verified in step S4 as follows: Known and And simultaneously satisfy and ,get : For closed-loop networked inspection robot dog cluster system If the following conditions are met: i) ;ii) ;iii) ; This ensures the recursive feasibility of the proposed control method.

8. The efficient intelligent networked inspection robot dog cluster collaborative control method according to claim 6, characterized in that, The stability of the closed-loop system is verified in step S4 as follows: Under the condition that recursion feasibility is satisfied, the inspection robot dog cluster system under adaptive time-domain event-triggered distributed model predictive control will remain stable if the following conditions are met: If the conditions are met ,in ,as well as Then, the closed-loop system will remain in the terminal domain after entering the terminal domain.

Images