Inertial navigation AGV motion control method based on dynamic coupling and self-adaption

Abstract

The invention relates to the technical field of robot control, in particular to an inertial navigation AGV motion control method based on dynamic coupling and self-adaption. Performing dynamic slip rule analysis on the AGV through the wheel slip monitoring data and the wheel-ground acting force monitoring data to obtain a dynamic evolution rule of wheel slip caused by steering torque when the AGV runs; constructing a vehicle dynamics model of the AGV vehicle, and performing dynamics coupling analysis of wheel slippage on the vehicle dynamics model based on a dynamic evolution rule to obtain the instability degree of the wheel slippage of the AGV vehicle to the vehicle body stability; and based on the Lyapunov theory, inversion recursion of a non-linear uncertainty adaptive servo system is carried out to construct an adaptive law, control gain is dynamically adjusted through the adaptive law, and a stable control strategy of AGV vehicle execution transportation demands is obtained. According to the invention, the AGV vehicle can realize high-precision motion control in a multi-obstacle environment, and the operation efficiency of the AGV vehicle in a complex dynamic environment is improved.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a motion control method for inertial navigation AGVs based on dynamic coupling and adaptation. Background Technology

[0002] Automated Guided Vehicles (AGVs) are intelligent transportation devices capable of autonomously navigating along predetermined paths, widely used in industrial automation, warehousing and logistics, and intelligent manufacturing. Inertial navigation AGVs, in particular, utilize sensors such as inertial measurement units and odometers to acquire their attitude and position, combining this with navigation algorithms to achieve autonomous positioning and path tracking. They offer advantages such as independence from external signals and strong environmental adaptability, demonstrating significant strengths in complex environments. However, the motion characteristics of AGVs are affected by the dynamic coupling between the vehicle body and wheels. Especially during steering, acceleration, or driving on complex road surfaces, tire sideslip, slippage, and load changes can cause a mismatch between control commands and actual motion responses, reducing the accuracy and stability of motion control. Existing methods typically employ control strategies based on simplified models, failing to fully consider the dynamic coupling effects, thus limiting control performance.

[0003] Secondly, existing AGV motion control algorithms mostly employ fixed parameter control, but these methods struggle to adapt to changes in load, terrain, and environment in real time, leading to performance degradation in complex scenarios. The lack of effective adaptive mechanisms limits the robustness and adaptability of AGVs in dynamic environments. Furthermore, due to the cumulative error problem in inertial navigation, AGVs experience position drift after prolonged operation, resulting in significant deviations in the system's tracking of motion trajectories and path planning. This increases the likelihood of collisions with obstacles, thus affecting the navigation accuracy of the AGV vehicles. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides a motion control method for inertial navigation AGVs based on dynamic coupling and adaptation.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a motion control method for inertial navigation AGVs based on dynamic coupling and adaptation, comprising the following steps: S102: By analyzing the slip law dynamics of AGV vehicles through wheel slip monitoring data and wheel-ground force monitoring data, the dynamic evolution law of wheel slip caused by steering torque during AGV vehicle operation is obtained. S104: Construct a vehicle dynamics model for the AGV vehicle, and perform dynamic coupling analysis of wheel slippage on the vehicle dynamics model based on the dynamic evolution law to obtain the degree of instability of wheel slippage of the AGV vehicle to the stability of the vehicle body. S106: An adaptive servo system based on Lyapunov theory inversion and recursive nonlinear uncertainty is used to construct an adaptive law. The control gain is dynamically adjusted through the adaptive law to obtain a stable control strategy for AGV vehicles to meet transportation needs. S108: Based on vehicle dynamics and nonlinear characteristics, the stability control strategy is adjusted in real time according to the degree of instability to optimize the motion trajectory and obtain the optimal obstacle avoidance and tracking trajectory of the AGV vehicle. S110: Based on the optimal obstacle avoidance and tracking trajectory, randomly expand the path points of the AGV vehicle in the transportation scenario to generate an expanded random tree. Based on the expanded random tree, improve the planned optimal path to control the inertial navigation motion of the AGV vehicle.

[0006] More specifically, step S102 includes the following steps: Acquire the preset in-place turning strategy of the AGV vehicle, and at the same time acquire the wheel slip monitoring data of the AGV vehicle executing the preset in-place turning strategy within a preset time period, as well as the wheel-ground force monitoring data when generating each wheel slip monitoring data. Using a preset time period as the time sequence reference, the monitoring data of each wheel's ground force is selected as the output of the execution action according to the preset in-situ steering strategy. Based on the time sequence reference, the outputs of each execution action and the corresponding wheel slip monitoring data are combined to construct the wheel slip force trajectory. The system presets wheel slip force factors and continuously accesses the steering torque of each wheel slip state-wheel-ground force action pair on the wheel slip force trajectory based on the wheel slip force factors. During the access process, the current wheel slip state is marked as an accessed state and the cumulative steering torque reward from the start of the wheel slip state to the end of the wheel slip force trajectory is calculated to obtain a series of steering torque reward values. If the currently visited wheel slip state is already visited, then the original value of the wheel slip state for the state and action on the wheel slip force trajectory is updated based on the average of a series of steering torque return values, thus obtaining the new value estimation function. Repeat the above steps of calculating the steering torque return value and updating the original value to access and iterate the remaining wheel slip state-wheel-ground force action pairs to generate a new value estimation matrix. Based on the new value estimation matrix, determine the dynamic evolution law of wheel slip caused by steering torque when the AGV is running.

[0007] More specifically, step S104 includes the following steps: Historical vehicle posture data and historical uneven load distribution data of each wheel were extracted from the operation logs when the AGV vehicle executed the preset in-situ turning strategy. By introducing nonlinear dynamics theory, a Newton-Euler equation for nonlinear dynamics is constructed. The linear momentum and angular momentum of the historical vehicle body attitude data and the historical non-uniform load distribution data of each wheel are solved by the Newton-Euler equation to determine the overturning coefficient of the non-uniform load distribution on the vehicle body attitude during the AGV vehicle turning process. Obtain the structural parameters of the AGV vehicle, construct a geometric simulation model of the AGV vehicle based on the structural parameters, and perform simulation analysis and simulation of the vehicle body attitude change and dynamic distribution of load on each wheel in the dynamic simulation software based on the overturning coefficient to generate the vehicle dynamic model of the AGV vehicle. The multibody dynamics solution method is introduced. Based on the dynamic evolution law, the coupling mechanism of wheel slippage with the vehicle dynamics model is dynamically calculated in the multibody dynamics solution method to obtain the coupling index. The degree of instability of wheel slippage of AGV vehicle with respect to vehicle stability is determined according to the coupling index.

[0008] More specifically, step S106 includes the following steps: The system obtains the transportation demand of AGV vehicles, acquires the current control strategy of the adaptive servo system to determine the transportation demand, simulates the current control strategy using the vehicle dynamics model of AGV vehicles, and outputs the simulation control parameter set and simulation control response index of the adaptive servo system. The desired control response index is preset, and a nonlinear hierarchical method is introduced to construct a set of nonlinear state equations for the adaptive servo control system. The error between the control response index and the desired control response index is calculated to obtain the control response error value. Based on the control response error value, the Lyapunov function is preset and the transformation is performed to obtain the Lyapunov derivation. Each equation in the nonlinear state equation set is substituted into the Lyapunov derivation and solved recursively to obtain several virtual inversion control quantities of the adaptive servo control system. If there is one or more uncertain parameters in the simulation control parameter set, the uncertain parameters in the simulation control parameter set are extracted and parameter error estimation is performed to obtain parameter estimation error terms. The parameter estimation error terms of each uncertain parameter are added to the Lyapunov derivation to obtain the modified Lyapunov function. An adaptive law is constructed based on the modified Lyapunov function. If the control response error value is greater than 0, the control gain is dynamically adjusted through an adaptive law until the control response error value is less than or equal to 0, generating adaptive gain compensation. Several virtual inversion control quantities are adjusted using adaptive gain compensation to obtain the final control input. Based on the final control input, a stable control strategy for the AGV vehicle to perform transportation needs is determined.

[0009] More specifically, step S108 includes the following steps: Obtain environmental perception information of the AGV vehicle in the target transportation scenario, obtain functional information of different obstacles based on the environmental perception information, and extract multiple historical duty routes of each obstacle from the transportation log based on the functional information. Convolutional neural networks are constructed, and multiple historical duty performance routes are trained and validated through convolutional neural networks to obtain a dynamic duty performance prediction model for obstacles. The dynamic duty performance prediction model for obstacles is used to predict the current obstacle in order to obtain the duty performance prediction route of the current obstacle at the next time node. The expected stability of AGV vehicle inertial navigation is extracted by transportation demand. Based on the final control input, a penalty term function is preset for the vehicle's dynamic constraints and nonlinear characteristics. The Lagrange algorithm is introduced, and a series of Lagrange functions for dynamic obstacle avoidance prediction are constructed in the Lagrange algorithm according to the predicted route. The vehicle motion iterative path is obtained by minimizing the weighted calculation of the expected stability and the penalty term function. The vehicle motion iterative path is continuously tracked by the Lagrangian function. During the tracking process, the iterative stability control strategy is updated based on the instability to approximate the optimal trajectory solution for the AGV vehicle inertial control to reach the expected stability. The current iteration step size is output. The iteration step size threshold is preset according to the expected stability. If the current iteration step size does not exceed the iteration step size threshold, the tracking iteration continues until it does not exceed the iteration step size threshold, and then the tracking iteration operation stops, generating the optimal obstacle avoidance and tracking trajectory of the AGV vehicle in a multi-obstacle environment.

[0010] More specifically, step S110 includes the following steps: Construct a path search space for the movement of AGV vehicles within the target transportation scenario, obtain the transportation start point and transportation end point of the AGV vehicles, and construct an extended random tree architecture for the path search space. Based on the transportation origin, a random path point is sampled in the path search space, and the nearest point whose Euclidean distance to the random path point is less than a preset Euclidean distance is found and defined as the adjacent path point. The expansion step size range is set according to the optimal obstacle avoidance tracking trajectory. If the random path point is not within the expansion step size range, the expansion starts from the adjacent path point towards the random path point within the maximum step size limit. If the random path point is within the expansion step size range, the random path point is directly extracted to generate a new random path node. An obstacle pattern distribution map is constructed based on environmental perception information. The obstacle pattern distribution map is embedded into the path search space. It is then checked whether the path from the adjacent path point to the newly generated random path node overlaps or collides with the obstacle pattern distribution map. If there is no overlap or collision, the newly generated random path node is added to the extended random tree structure and connected to the adjacent path node. If an overlap or collision occurs, the newly generated random path node is discarded and the extension is resampled until the Euclidean distance between the newly generated random path node and the transportation destination is greater than the preset Euclidean distance, and the extended random tree of the path planning is generated. Extract the leaf node values ​​and growth layout of each leaf node value from the extended random tree. Based on the growth layout, backtrack each leaf node value from the transportation destination to obtain the optimal planning path. Control the inertial navigation motion of the AGV vehicle according to the optimal planning path.

[0011] This invention addresses the technical deficiencies in the prior art, and its beneficial technical effects are as follows: By analyzing the slip dynamics of AGV vehicles using wheel slip monitoring data and wheel-ground force monitoring data, the dynamic evolution law of wheel slip caused by steering torque during AGV operation is obtained. A vehicle dynamics model of the AGV vehicle is constructed, and dynamic coupling analysis of wheel slip is performed on the vehicle dynamics model based on the dynamic evolution law to obtain the degree of instability of wheel slip on the vehicle body stability. Based on Lyapunov theory, an adaptive servo system with recursive nonlinear uncertainty is derived to construct an adaptive law. The control gain is dynamically adjusted through the adaptive law to obtain a stable control strategy for the AGV vehicle to perform transportation needs. With vehicle dynamics and nonlinear characteristics as constraints, the control input of the stable control strategy for the predicted route is adjusted in real time based on the degree of instability to optimize the motion trajectory and obtain the optimal obstacle avoidance and tracking trajectory of the AGV vehicle. Based on the optimal obstacle avoidance and tracking trajectory, the path points of the AGV vehicle in the transportation scenario are randomly expanded to generate an extended random tree. The optimal path planning of the AGV vehicle is improved based on the extended random tree to control the inertial navigation transportation of the AGV vehicle. This invention enables motion control of AGV vehicles by analyzing the coupling mechanism between wheel slip and vehicle dynamics and Lyapunov's adaptive control optimization, thereby achieving high-precision motion control in multi-obstacle environments, ensuring the stability and robustness of the system under external disturbances, and improving the operating efficiency of AGV vehicles in complex dynamic environments. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0013] Figure 1A flowchart of the first method for motion control of inertial navigation AGV based on dynamic coupling and adaptation is shown. Figure 2 A flowchart of the second method for motion control of inertial navigation AGV based on dynamic coupling and adaptation is shown; Figure 3 A flowchart of the third method for motion control of inertial navigation AGVs based on dynamic coupling and adaptation is shown. Detailed Implementation

[0014] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0015] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0016] The first aspect of this invention provides a motion control method for inertial navigation AGVs based on dynamic coupling and adaptation, such as... Figure 1 As shown, it includes the following steps: S102: By analyzing the slip law dynamics of AGV vehicles through wheel slip monitoring data and wheel-ground force monitoring data, the dynamic evolution law of wheel slip caused by steering torque during AGV vehicle operation is obtained. S104: Construct a vehicle dynamics model for the AGV vehicle, and perform dynamic coupling analysis of wheel slippage on the vehicle dynamics model based on the dynamic evolution law to obtain the degree of instability of wheel slippage of the AGV vehicle to the stability of the vehicle body. S106: An adaptive servo system based on Lyapunov theory inversion and recursive nonlinear uncertainty is used to construct an adaptive law. The control gain is dynamically adjusted through the adaptive law to obtain a stable control strategy for AGV vehicles to meet transportation needs. S108: Based on vehicle dynamics and nonlinear characteristics, the stability control strategy is adjusted in real time according to the degree of instability to optimize the motion trajectory and obtain the optimal obstacle avoidance and tracking trajectory of the AGV vehicle. S110: Based on the optimal obstacle avoidance and tracking trajectory, randomly expand the path points of the AGV vehicle in the transportation scenario to generate an expanded random tree. Based on the expanded random tree, improve the planned optimal path control AGV vehicle inertial navigation transportation.

[0017] More specifically, step S102, as follows: Figure 2As shown, the specific steps include: S202: Obtain the preset in-place turning strategy of the AGV vehicle, and at the same time obtain the wheel slip monitoring data of the AGV vehicle executing the preset in-place turning strategy within a preset time period, as well as the wheel-ground force monitoring data when generating each wheel slip monitoring data. S204: Using a preset time period as the time sequence reference, select the monitoring data of each wheel's ground force as the output of the execution action according to the preset in-situ steering strategy, and construct the wheel slip force trajectory by combining each execution action output and the corresponding wheel slip monitoring data based on the time sequence reference. S206: Preset wheel slip force factors, continuously access the steering torque of each wheel slip state-wheel-ground force action pair on the wheel slip force trajectory based on the wheel slip force factors, mark the current wheel slip state as the accessed state during the access process and calculate the cumulative steering torque reward from the start of the wheel slip state to the end of the wheel slip force trajectory to obtain a series of steering torque reward values. S208: If the currently visited wheel slip state is already visited, then update the original value of the wheel slip state for the state and action on the wheel slip force trajectory based on the average of a series of steering torque return values, and obtain the new value estimation function. S210: Repeat the above steps of calculating the steering torque return value and updating the original value to access and iterate the remaining wheel slip state-wheel-ground force action pairs to generate a new value estimation matrix. Based on the new value estimation matrix, determine the dynamic evolution law of wheel slip caused by steering torque when the AGV is running.

[0018] It should be noted that traditional models in current dynamic studies of AGV vehicle in-situ turning are usually based on idealized assumptions, such as ignoring wheel slip and uniform ground pressure distribution. However, such simplified models cannot accurately describe the interaction between the tires and the ground under complex road conditions, leading to inaccurate predictions of the dynamic performance of AGV vehicles during in-situ turning. Especially under non-homogeneous ground or high load conditions, the ambiguity of tire slip and steering torque can significantly affect the steering stability of AGVs, thereby reducing the control accuracy and motion reliability of AGV vehicles. Therefore, revealing the evolution law of steering torque and wheel slip during AGV vehicle operation is crucial. To address this, this method uses wheel slip monitoring data and corresponding wheel-ground interaction force monitoring data to construct a time-series wheel slip force trajectory. This wheel slip force trajectory is a complete cycle of strategy execution, reflecting the existing empirical accumulation of the relationship between wheel slip and steering torque. This ensures that the evolution law of wheel slip and steering torque is based on the interaction with the actual environment rather than relying on an environmental model, thus improving the reliability of the law deduction and analysis. Since the steering torque of an AGV vehicle when turning in place may generate different wheel-ground interaction forces, it is necessary to focus on factors such as wheel slippage and uneven distribution of ground pressure. That is, based on the wheel slippage force factor, we continuously visit each set of wheel slippage states and the corresponding wheel-ground interaction force actions of the steering torque on the wheel slippage force trajectory. In this way, we can explore the effect of wheel slippage on steering torque reward under the complete strategy execution round. During the visit, we calculate the cumulative steering torque reward from the beginning of the wheel slippage state to the end of the wheel slippage force trajectory. Based on the reward calculation condition of the complete round, we can correctly evaluate the contribution of wheel slippage to the long-term steering torque reward, determine the update direction of the evolution law, and make the deduction of the evolution law of steering torque causing wheel slippage more accurate and reasonable.

[0019] It should be noted that for each visited wheel slip state, the initial value of the wheel slip state and the corresponding steering torque action is updated using the average of a series of obtained steering torque reward values. The update of the wheel slip state is performed using the average of all rewards when the state is visited, while the update of the steering torque action is completed using the average of all rewards when the state is visited and the action is performed accordingly.

[0020] This allows us to reveal the correct weighting rules of the state generated by the action through value updates, enabling the state deduction and estimation of wheel slippage to converge to the action reward of the actual steering torque. The main formula for the update is: ; in, For the newborn value estimation function, For original value, For learning rate, This is the return value for steering torque.

[0021] Finally, by iteratively accessing the remaining wheel slip states and wheel-ground force actions, the newly generated value estimation matrix can quickly and accurately clarify the dynamic evolution law of wheel slip caused by steering torque during AGV vehicle operation. This method can explore the evolution law of tire slip and steering torque under complex road conditions, analyze its influence mechanism on steering stability, and provide a reliable analytical foundation for improving the steering stability and motion control capability of AGV vehicles under complex working conditions.

[0022] More specifically, step S104 includes the following steps: Historical vehicle posture data and historical uneven load distribution data of each wheel were extracted from the operation logs when the AGV vehicle executed the preset in-situ turning strategy. By introducing nonlinear dynamics theory, a Newton-Euler equation for nonlinear dynamics is constructed. The linear momentum and angular momentum of the historical vehicle body attitude data and the historical non-uniform load distribution data of each wheel are solved by the Newton-Euler equation to determine the overturning coefficient of the non-uniform load distribution on the vehicle body attitude during the AGV vehicle turning process. Obtain the structural parameters of the AGV vehicle, construct a geometric simulation model of the AGV vehicle based on the structural parameters, and perform simulation analysis and simulation of the vehicle body attitude change and dynamic distribution of load on each wheel in the dynamic simulation software based on the overturning coefficient to generate the vehicle dynamic model of the AGV vehicle. The multibody dynamics solution method is introduced. Based on the dynamic evolution law, the coupling mechanism of wheel slippage with the vehicle dynamics model is dynamically calculated in the multibody dynamics solution method to obtain the coupling index. The degree of instability of wheel slippage of AGV vehicle with respect to vehicle stability is determined according to the coupling index.

[0023] It should be noted that AGVs often face changes in vehicle posture and uneven load distribution when turning in place. This uneven load can lead to wheel slippage, abnormal tire wear, and even affect the overall stability of the vehicle. Traditional methods often focus on static load analysis or simplified dynamic models, failing to fully consider the influence of vehicle structural parameters on dynamic response, and thus struggling to accurately describe the dynamic stability mechanism of AGVs under complex working conditions with geometric deformation and load distribution changes. Secondly, there is a mechanism of coupling between dynamics and wheel slippage during AGV turning in place, but existing methods have limitations in modeling the mechanical coupling of wheel slippage, failing to fully reveal the degree of instability during AGV turning. Therefore, this method constructs the Newton-Euler equations for nonlinear dynamics to solve for the historical vehicle posture data and historical non-uniform load distribution data of AGV vehicles using nonlinear dynamics theory. The Newton-Euler equations are a recursive method for rigid body dynamics analysis, used to calculate the force and motion states of multi-rigid-body vehicle systems. First, the linear and angular accelerations of each moving rigid body of the vehicle are calculated based on the historical vehicle posture data and historical non-uniform load distribution data. Then, the Newtonian equations of motion (linear momentum) and the Euler equations of motion (angular momentum) are calculated based on the linear and angular accelerations. After the calculations are completed, the vehicle dynamics and torques are calculated recursively from the end rigid body to the root rigid body. Thus, the overturning coefficient of the non-uniform load distribution on the vehicle posture during the AGV vehicle's turning process can be determined based on the vehicle dynamics and torques. The overturning coefficient measures the stability of the wheels due to the geometric deformation of the vehicle body under changes in load distribution. The formula for the Newtonian equations of motion is: ; in, Let the mass of the rigid body be... Let be the linear acceleration of the center of mass of the rigid body, in units of . , It is the resultant force of all external forces acting on a rigid body.

[0024] The formula for Euler's equations of motion is as follows: ; in, Let the inertia tensor of the rigid body be... Let be the angular acceleration of the rigid body. Let ω be the angular velocity of the rigid body. The centrifugal torque of the AGV vehicle relative to its wheels. It is the resultant force of all external torques acting on the rigid body.

[0025] It should be noted that the dynamic calculation of the coupling mechanism between wheel slip and the vehicle dynamics model in the multibody dynamics solution method based on dynamic evolution laws specifically includes: establishing a contact mechanics model, calculating the contact forces according to the dynamic evolution laws through the contact mechanics model, including the normal force (elastic restoring force) and tangential force (friction force) applied by the vehicle dynamics to the wheel slip, then constructing the dynamic equations of the contact between the vehicle dynamics and the wheel slip, and solving these dynamic equations using the calculated contact forces to obtain a series of data such as the trajectory, velocity, and contact stress between the AGV vehicle and the wheel slip. Based on these data, further mechanical analysis can be performed to determine the coupling mechanism between wheel slip and the vehicle dynamics model, thereby generating the coupling degree of the contact interaction between the two, i.e., the coupling index. Based on the coupling index, the degree of instability of the AGV vehicle's wheel slip on the stability of the vehicle body can be determined. The dynamic equations satisfy the following relationship: ; Where M is the mass matrix and q is the generalized coordinate. As an external force, It is the contact force.

[0026] More specifically, step S106, as follows: Figure 3 As shown, the specific steps include: S302: Obtain the transportation demand of the AGV vehicle, obtain the current control strategy of the adaptive servo system to decide the transportation demand, simulate the current control strategy through the vehicle dynamics model of the AGV vehicle, and output the simulation control parameter set and simulation control response index of the adaptive servo system. S304: Preset the desired control response index, introduce a nonlinear hierarchical method to construct a nonlinear state equation set for the adaptive servo control system, calculate the error between the control response index and the desired control response index, and obtain the control response error value. S306: Based on the control response error value, preset the Lyapunov function and perform transformation and differentiation to obtain the Lyapunov derivation. Substitute each equation in the nonlinear state equation set into the Lyapunov derivation and solve it recursively to obtain several virtual inversion control quantities of the adaptive servo control system. S308: If there is one or more uncertain parameters in the simulation control parameter set, extract the uncertain parameters in the simulation control parameter set and perform parameter error estimation to obtain parameter estimation error terms. Add the parameter estimation error terms of each uncertain parameter to the Lyapunov derivation to obtain the modified Lyapunov function. Construct an adaptive law based on the modified Lyapunov function. S310: If the control response error value is greater than 0, the control gain is dynamically adjusted through an adaptive law until the control response error value is less than or equal to 0, generating adaptive gain compensation. Several virtual inversion control quantities are adjusted using adaptive gain compensation to obtain the final control input. Based on the final control input, a stable control strategy for the AGV vehicle to perform transportation needs is determined.

[0027] It should be noted that existing adaptive servo control systems for AGV vehicles face insufficient control accuracy in complex dynamic environments. This is primarily due to the uncertainty of system parameters and the influence of external disturbances, making it difficult for traditional control methods to adjust parameters in real time and ensure system stability. Secondly, traditional control methods have significant limitations in handling nonlinear and uncertain problems, struggling to effectively cope with parameter changes and external disturbances during servo system operation. Especially under complex operating conditions, fixed-gain control strategies struggle to balance system stability and response performance, potentially leading to decreased control accuracy or even system oscillations or instability. To address this, this method first simulates the current control strategy of the adaptive servo system in determining transportation needs using a vehicle dynamics model of the AGV, thereby clarifying the control parameters for the current AGV vehicle motion and the system's own control response performance. Next, inverse control is performed on the system to assess response performance. The nonlinear state equations are described using a strict feedback nonlinear form, clearly defining the role of the final control input in the adaptive servo system and ensuring the control scheme is applicable to nonlinear systems. The control response error value, compared to the expected control response index, actually reflects the error variable of the adaptive servo system. This error variable can be used to gradually decouple the system, and its introduction allows the system to gradually stabilize, making inversion control more feasible. Then, based on Lyapunov theory, each equation of the nonlinear state equation set is recursively derived according to the control response error value. This step involves inverting and recursively deriving the adaptive servo system based on the error variable. By recursively designing Lyapunov functions to reduce the error of certain state variables, the system response selects appropriate virtual inversion control quantities to gradually stabilize each error state. This improves the system's response performance in handling nonlinear uncertainties and enhances the accuracy of the system's operational control decisions for AGV vehicles.

[0028] It should be noted that if one or more uncertain parameters exist in the simulated control parameter set, it indicates that the current adaptive servo system is affected by external disturbances and parameter changes during the control decision-making process, significantly reducing its control stability and performance. Therefore, this method designs an adaptive law adapted to the system to dynamically adjust the control gain of the AGV motion, overcoming the uncertainty of the system model parameters and ensuring the robustness and reliability of the servo system under complex operating conditions. Specifically, in designing the adaptive law, this method accurately estimates the parameter estimation error term of the uncertain parameters, transforming the estimation problem of uncertain parameters into a problem of controlling system errors. This allows the designed adaptive law to better adjust the system parameter estimates, providing a reference for eliminating uncertain parameters. Then, the Lyapunov function is modified using the parameter estimation error terms of each uncertain parameter, enabling the adaptive law design to simultaneously consider the stability of the system state and the convergence of parameter estimation. If the control response error is greater than 0, it indicates that the system has failed to converge quickly and maintain stability under external disturbances and parameter changes. Therefore, control gain compensation adjustment is required. The constructed adaptive law can be used to further obtain the control gain of the dynamically adjusted system to address parameter uncertainties. Finally, the control gain is used back to the virtual inversion control quantity, resulting in a more stable AGV vehicle adaptive control strategy. This method introduces an inversion control strategy and combines it with Lyapunov theory to construct an adaptive servo control model, achieving dynamic adjustment of the control gain. This enhances the system's adaptability to uncertainties and improves the robustness and response performance of the servo system in complex environments. Simultaneously, based on Lyapunov stability theory, a nonlinear dynamic adaptive law is designed to achieve online real-time adjustment of system parameters, improving the system's anti-interference capability and stability when making AGV vehicle control decisions.

[0029] More specifically, step S108 includes the following steps: Obtain environmental perception information of the AGV vehicle in the target transportation scenario, obtain functional information of different obstacles based on the environmental perception information, and extract multiple historical duty routes of each obstacle from the transportation log based on the functional information. Convolutional neural networks are constructed, and multiple historical duty performance routes are trained and validated through convolutional neural networks to obtain a dynamic duty performance prediction model for obstacles. The dynamic duty performance prediction model for obstacles is used to predict the current obstacle in order to obtain the duty performance prediction route of the current obstacle at the next time node. The expected stability of AGV vehicle inertial navigation is extracted by transportation demand. Based on the final control input, a penalty term function is preset for the vehicle's dynamic constraints and nonlinear characteristics. The Lagrange algorithm is introduced, and a series of Lagrange functions for dynamic obstacle avoidance prediction are constructed in the Lagrange algorithm according to the predicted route. The vehicle motion iterative path is obtained by minimizing the weighted calculation of the expected stability and the penalty term function. The vehicle motion iterative path is continuously tracked by the Lagrangian function. During the tracking process, the iterative stability control strategy is updated based on the instability to approximate the optimal trajectory solution for the AGV vehicle inertial control to reach the expected stability. The current iteration step size is output. The iteration step size threshold is preset according to the expected stability. If the current iteration step size does not exceed the iteration step size threshold, the tracking iteration continues until it does not exceed the iteration step size threshold, and then the tracking iteration operation stops, generating the optimal obstacle avoidance and tracking trajectory of the AGV vehicle in a multi-obstacle environment.

[0030] It should be noted that current AGV vehicle trajectory tracking accuracy is relatively low in complex environments with wheel slippage dynamics coupling. Traditional trajectory tracking methods often fail to adequately consider vehicle dynamic constraints and nonlinear characteristics when facing dynamic environments, unstructured roads, or multiple obstacles, leading to decreased path planning accuracy and large tracking errors, thus affecting the operational stability and efficiency of AGVs. Furthermore, existing obstacle avoidance strategies often rely on preset rules or simple path replanning, making it difficult to balance real-time performance and global optimality, resulting in poor obstacle avoidance performance in complex environments. To address this, our method constructs a dynamic obstacle performance prediction model using multiple historical performance routes of different obstacles performing their respective functions. This model predicts the performance route of the current obstacle at the next time point. The obstacle route prediction model provides accurate and reliable obstacle avoidance targets for subsequent AGV vehicle operation, significantly improving the optimization accuracy of trajectory tracking compared to traditional methods. Then, under the constraints of vehicle dynamics and nonlinear characteristics, this method performs optimal trajectory tracking with the desired stability based on the instability of in-situ turning in dynamic obstacle avoidance control of AGV vehicles. This enables AGV vehicles to fully consider vehicle dynamics and nonlinear characteristics constraints under the premise of in-situ turning stability, optimize control input, and further rationally plan obstacle avoidance tracking trajectories for future obstacle movement changes. It effectively combines the dynamic obstacle avoidance mechanism provided by environmental perception information to reduce tracking errors, enabling AGV vehicles to achieve stable and efficient trajectory tracking in multi-obstacle environments, thereby improving the system's intelligence level and adaptability.

[0031] It should be noted that the desired stability level is a threshold value for the AGV to achieve a stable turning performance in place. It is the objective function for maintaining motion stability during the AGV's dynamic obstacle avoidance trajectory tracking. Since the final control input decision for the AGV's movement is constrained by the vehicle's own dynamics and nonlinear characteristics, a constraint penalty condition, i.e., a penalty term function, needs to be set based on these characteristics. This ensures that the trajectory tracking optimization process strictly adheres to the criteria of vehicle dynamics and nonlinear characteristics. The Lagrangian function incorporates the AGV's dynamic constraints into the objective function. By optimizing this Lagrangian function, the optimal solution for trajectory tracking is gradually approximated, allowing for richer solutions and improving the selectivity and constraint of trajectory optimization. Next, a path within the feasible region of the constraints is defined based on the weighted sum and minimization of the objective function and the penalty term function. This path clarifies the iterative direction in which the AGV vehicle can maximize its avoidance of future obstacles while achieving the desired stability under the constraints. By tracking this path through obstacle avoidance control, the optimal solution can be approximated. Simultaneously, since this path lies within the feasible region of the constraints, it avoids touching the constraint boundaries during the tracking optimization process. The iteration step size determines the amount of solution update in each iteration. In each iteration, the step size needs to be adjusted based on the current solution and changes in the objective function to balance convergence speed and stability, enabling a faster finding of the optimal trajectory solution.

[0032] More specifically, step S110 includes the following steps: Construct a path search space for the movement of AGV vehicles within the target transportation scenario, obtain the transportation start point and transportation end point of the AGV vehicles, and construct an extended random tree architecture for the path search space. Based on the transportation origin, a random path point is sampled in the path search space, and the nearest point whose Euclidean distance to the random path point is less than a preset Euclidean distance is found and defined as the adjacent path point. The expansion step size range is set according to the optimal obstacle avoidance tracking trajectory. If the random path point is not within the expansion step size range, the expansion starts from the adjacent path point towards the random path point within the maximum step size limit. If the random path point is within the expansion step size range, the random path point is directly extracted to generate a new random path node. An obstacle pattern distribution map is constructed based on environmental perception information. The obstacle pattern distribution map is embedded into the path search space. It is then checked whether the path from the adjacent path point to the newly generated random path node overlaps or collides with the obstacle pattern distribution map. If there is no overlap or collision, the newly generated random path node is added to the extended random tree structure and connected to the adjacent path node. If an overlap or collision occurs, the newly generated random path node is discarded and the extension is resampled until the Euclidean distance between the newly generated random path node and the transportation destination is greater than the preset Euclidean distance, and the extended random tree of the path planning is generated. Extract the leaf node values ​​and growth layout of each leaf node value from the extended random tree. Based on the growth layout, backtrack each leaf node value from the transportation destination to obtain the optimal planning path. Control the inertial navigation motion of the AGV vehicle according to the optimal planning path.

[0033] It should be noted that existing AGV path planning methods often suffer from shortcomings in complex environments, such as insufficient adaptability to dynamic obstacles, significant susceptibility of path planning quality to environmental changes, and limited local path adjustment capabilities. While traditional methods can generate feasible paths, these paths may contain redundancy in complex or dynamic environments, affecting execution efficiency and path smoothness, and making real-time path optimization difficult. Therefore, this method constructs a path search space and a corresponding extended random tree architecture to enable the adaptive servo system to explore motion paths under obstacle avoidance conditions. To make the path search from the starting point to the destination more efficient, a path point is randomly sampled as the exploration fulcrum covered by the system. The nearest node to this random path point, i.e., the adjacent path point, is found to ensure that the expansion grows from the existing tree to the new area, guaranteeing the formation of a coherent path for the AGV. Since the planned optimal obstacle avoidance tracking trajectory is an accurate and fixed motion benchmark, this method sets the expansion step size range of the path point orientation based on the optimal obstacle avoidance tracking trajectory. This ensures that the path search follows the optimal obstacle avoidance tracking trajectory as a benchmark, guaranteeing that the AGV can achieve dynamic obstacle avoidance for the searched path. If a random pathpoint is not within the expansion step size range, it indicates that the planning of the random pathpoint is too dense, which may lead to dynamic obstacle avoidance errors and cause the AGV vehicle to collide with obstacles. Therefore, the expansion starts from the adjacent pathpoints towards the random pathpoint within the maximum step size limit. Conversely, if the random pathpoint is within the range, it indicates that the exploration of the random pathpoint is reasonable, and therefore the random pathpoint is directly extracted.

[0034] It should be noted that if the path from the adjacent path point to the newly generated random path node does not overlap or collide with the obstacle distribution map, it indicates that the expanded random exploration path can avoid the dynamic behavior of obstacles, representing a high feasibility of the path. If overlap or collision occurs, it indicates that the path has an expansion density error, causing slight deviations when the path faces changes in dynamic obstacles. Therefore, the newly generated random path node needs to be discarded and resampled and expanded until the Euclidean distance between the newly generated random path node and the transportation destination is greater than the preset Euclidean distance. Finally, an expanded random tree for path planning is generated to reflect the path exploration cluster in response to obstacle distribution and road conditions. The values ​​and growth layout of the leaf nodes on this expanded random tree represent the optimal path planning. Backtracking to the destination makes path extraction smoother. This method optimizes the path search process based on a fast expanded random tree, which not only improves the optimization effect of global path planning but also allows for dynamic adjustments when the local environment changes, improving the adaptability and robustness of AGVs in complex environments. On the other hand, it can effectively cope with different obstacle distributions and road conditions, improving the reliability and execution efficiency of path planning.

[0035] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An AGV motion control method based on dynamics coupling and self-adaption, characterized in that, Includes the following steps: S102: By analyzing the slip law dynamics of AGV vehicles through wheel slip monitoring data and wheel-ground interaction force monitoring data, the dynamic evolution law of wheel slip caused by steering torque during AGV vehicle operation is obtained. S104: Construct a vehicle dynamics model for the AGV vehicle, and perform dynamic coupling analysis of wheel slippage on the vehicle dynamics model based on the dynamic evolution law to obtain the degree of instability of wheel slippage of the AGV vehicle to the stability of the vehicle body. S106: An adaptive servo system based on Lyapunov theory inversion and recursive nonlinear uncertainty is used to construct an adaptive law. The control gain is dynamically adjusted through the adaptive law to obtain a stable control strategy for AGV vehicles to meet transportation needs. S108: Based on vehicle dynamics and nonlinear characteristics, the stability control strategy is adjusted in real time according to the degree of instability to optimize the motion trajectory and obtain the optimal obstacle avoidance and tracking trajectory of the AGV vehicle. S110: Based on the optimal obstacle avoidance and tracking trajectory, randomly expand the path points of the AGV vehicle in the transportation scenario to generate an expanded random tree. Based on the expanded random tree, improve the planned optimal path to control the inertial navigation motion of the AGV vehicle.

2. The AGV motion control method based on dynamics coupling and adaption according to claim 1, characterized in that, Step S102 specifically includes the following steps: Acquire the preset in-place turning strategy of the AGV vehicle, and at the same time acquire the wheel slip monitoring data of the AGV vehicle executing the preset in-place turning strategy within a preset time period, as well as the wheel-ground force monitoring data when generating each wheel slip monitoring data. Using a preset time period as the time sequence reference, the monitoring data of each wheel's ground force is selected as the output of the execution action according to the preset in-situ steering strategy. Based on the time sequence reference, the outputs of each execution action and the corresponding wheel slip monitoring data are combined to construct the wheel slip force trajectory. The system presets wheel slip force factors and continuously accesses the steering torque of each wheel slip state-wheel-ground force action pair on the wheel slip force trajectory based on the wheel slip force factors. During the access process, the current wheel slip state is marked as an accessed state and the cumulative steering torque reward from the start of the wheel slip state to the end of the wheel slip force trajectory is calculated to obtain a series of steering torque reward values. If the currently visited wheel slip state is already visited, then the original value of the wheel slip state for the state and action on the wheel slip force trajectory is updated based on the average of a series of steering torque return values, thus obtaining the new value estimation function. Repeat the above steps of calculating the steering torque return value and updating the original value to access and iterate the remaining wheel slip state-wheel-ground force action pairs to generate a new value estimation matrix. Based on the new value estimation matrix, determine the dynamic evolution law of wheel slip caused by steering torque when the AGV is running.

3. The AGV motion control method based on dynamics coupling and adaption according to claim 1, characterized in that, Step S104 specifically includes the following steps: Historical vehicle posture data and historical uneven load distribution data of each wheel were extracted from the operation logs when the AGV vehicle executed the preset in-situ turning strategy. By introducing nonlinear dynamics theory, a Newton-Euler equation for nonlinear dynamics is constructed. The linear momentum and angular momentum of the historical vehicle body attitude data and the historical non-uniform load distribution data of each wheel are solved by the Newton-Euler equation to determine the overturning coefficient of the non-uniform load distribution on the vehicle body attitude during the AGV vehicle turning process. Obtain the structural parameters of the AGV vehicle, construct a geometric simulation model of the AGV vehicle based on the structural parameters, and perform simulation analysis and simulation of the vehicle body attitude change and dynamic distribution of load on each wheel in the dynamic simulation software based on the overturning coefficient to generate the vehicle dynamic model of the AGV vehicle. The multibody dynamics solution method is introduced. Based on the dynamic evolution law, the coupling mechanism of wheel slippage with the vehicle dynamics model is dynamically calculated in the multibody dynamics solution method to obtain the coupling index. The degree of instability of wheel slippage of AGV vehicle with respect to vehicle stability is determined according to the coupling index.

4. The AGV motion control method based on dynamics coupling and adaption according to claim 1, characterized in that, Step S106 specifically includes the following steps: The system obtains the transportation demand of AGV vehicles, acquires the current control strategy of the adaptive servo system to determine the transportation demand, simulates the current control strategy using the vehicle dynamics model of AGV vehicles, and outputs the simulation control parameter set and simulation control response index of the adaptive servo system. The desired control response index is preset, and a nonlinear hierarchical method is introduced to construct a set of nonlinear state equations for the adaptive servo control system. The error between the control response index and the desired control response index is calculated to obtain the control response error value. Based on the control response error value, the Lyapunov function is preset and the transformation is performed to obtain the Lyapunov derivation. Each equation in the nonlinear state equation set is substituted into the Lyapunov derivation and solved recursively to obtain several virtual inversion control quantities of the adaptive servo control system. If there is one or more uncertain parameters in the simulation control parameter set, the uncertain parameters in the simulation control parameter set are extracted and parameter error estimation is performed to obtain parameter estimation error terms. The parameter estimation error terms of each uncertain parameter are added to the Lyapunov derivation to obtain the modified Lyapunov function. An adaptive law is constructed based on the modified Lyapunov function. If the control response error value is greater than 0, the control gain is dynamically adjusted through an adaptive law until the control response error value is less than or equal to 0, generating adaptive gain compensation. Several virtual inversion control quantities are adjusted using adaptive gain compensation to obtain the final control input. Based on the final control input, a stable control strategy for the AGV vehicle to perform transportation needs is determined.

5. The inertial navigation AGV motion control method based on dynamic coupling and adaptation according to claim 1, characterized in that, Step S108 specifically includes the following steps: Obtain environmental perception information of the AGV vehicle in the target transportation scenario, obtain functional information of different obstacles based on the environmental perception information, and extract multiple historical duty routes of each obstacle from the transportation log based on the functional information. Convolutional neural networks are constructed, and multiple historical duty performance routes are trained and validated through convolutional neural networks to obtain a dynamic duty performance prediction model for obstacles. The dynamic duty performance prediction model for obstacles is used to predict the current obstacle in order to obtain the duty performance prediction route of the current obstacle at the next time node. The expected stability of AGV vehicle inertial navigation is extracted by transportation demand. Based on the final control input, a penalty term function is preset for the vehicle's dynamic constraints and nonlinear characteristics. The Lagrange algorithm is introduced, and a series of Lagrange functions for dynamic obstacle avoidance prediction are constructed in the Lagrange algorithm according to the predicted route. The vehicle motion iterative path is obtained by minimizing the weighted calculation of the expected stability and the penalty term function. The vehicle motion iterative path is continuously tracked by the Lagrangian function. During the tracking process, the iterative stability control strategy is updated based on the instability to approximate the optimal trajectory solution for the AGV vehicle inertial control to reach the expected stability. The current iteration step size is output. The iteration step size threshold is preset according to the desired stability. If the current iteration step size does not exceed the iteration step size threshold, the tracking iteration continues until it does not exceed the iteration step size threshold, and then the tracking iteration operation stops, generating the optimal obstacle avoidance and tracking trajectory of the AGV vehicle in a multi-obstacle environment.

6. The inertial navigation AGV motion control method based on dynamic coupling and adaptation according to claim 1, characterized in that, Step S110 specifically includes the following steps: Construct a path search space for the movement of AGV vehicles within the target transportation scenario, obtain the transportation start point and transportation end point of the AGV vehicles, and construct an extended random tree architecture for the path search space. Based on the transportation origin, a random path point is sampled in the path search space, and the nearest point whose Euclidean distance to the random path point is less than a preset Euclidean distance is found and defined as the adjacent path point. The expansion step size range is set according to the optimal obstacle avoidance tracking trajectory. If the random path point is not within the expansion step size range, the expansion starts from the adjacent path point towards the random path point within the maximum step size limit. If the random path point is within the expansion step size range, the random path point is directly extracted to generate a new random path node. An obstacle pattern distribution map is constructed based on environmental perception information. The obstacle pattern distribution map is embedded into the path search space. It is then checked whether the path from the adjacent path point to the newly generated random path node overlaps or collides with the obstacle pattern distribution map. If there is no overlap or collision, the newly generated random path node is added to the extended random tree structure and connected to the adjacent path node. If an overlap or collision occurs, the newly generated random path node is discarded and the extension is resampled until the Euclidean distance between the newly generated random path node and the transportation destination is greater than the preset Euclidean distance, and the extended random tree of the path planning is generated. Extract the leaf node values ​​and growth layout of each leaf node value from the extended random tree. Based on the growth layout, backtrack each leaf node value from the transportation destination to obtain the optimal planned path. Control the inertial navigation motion of the AGV vehicle according to the optimal planned path.

Images