Wheelchair dynamic modeling method based on load perception and slip compensation

Abstract

The invention discloses a wheelchair dynamic modeling method based on load sensing and slip compensation, which comprises the following steps: determining the total mass and the mass center position of a wheelchair based on acquired multi-source sensing data, and calculating the normal force of each hub; establishing an effective radius model and a rigidity model of the driving wheel hub; fusing slip compensation to establish a kinematic model of the wheelchair so as to calculate the linear velocity and the rotation angular velocity of the wheelchair; constructing a to-be-identified parameter vector; carrying out online identification and updating on the to-be-identified parameter vector based on a recursive least square method; correcting the output of the kinematic model by using a parameter vector; and outputting the model for pose estimation and motion control of the wheelchair. According to the method, the load and motion states are sensed in real time through multi-source sensing fusion, and the wheelchair dynamic model fusing the load and the slip effect is established and corrected online, so that the motion control precision and adaptability of the wheelchair under the variable load and complex road surface are improved.

Description

Technical Field

[0001] This invention relates to the field of dynamics modeling technology, and in particular to a wheelchair dynamics modeling method based on load sensing and slip compensation. Background Technology

[0002] In recent years, with the increasing aging population and growing demand for rehabilitation, intelligent wheelchairs and integrated bed-chair devices have been widely used in medical rehabilitation and home care. Intelligent wheelchairs typically employ rear-wheel differential drive, achieving steering and straight-line movement by adjusting the speed of the left and right drive wheels, thus simplifying the mechanical structure. This is one of the most common drive methods. Integrated bed-chair care robots, on the other hand, are innovative products designed to address the care needs of an aging society and people with disabilities. They enable users to adjust between beds and chairs in multiple positions, move from the bed to a wheelchair for easier mobility, and are easily used at home.

[0003] Existing technologies have yielded some research on the control and motion models of integrated wheelchair and bed-chair devices. One approach is the ideal differential kinematic model, which assumes no slippage and constant geometric parameters, using the conversion between linear and angular velocities for path control. Another approach is the extended dynamic model, which attempts to improve control accuracy by considering factors such as mass, moment of inertia, and friction; however, the parameters in these models are mostly empirical or static values. Furthermore, another category of research focuses on slip detection and control compensation. While these methods can detect and correct slippage at the control level, they rarely model changes in the center of mass and normal force caused by loads as input parameters.

[0004] These existing technical solutions each have some shortcomings: traditional models do not fully consider the influence of seat pressure and the center of pressure on the center of gravity offset, thus failing to accurately reflect the dynamic distribution of normal forces on the left and right and front and rear wheels of the wheelchair. Insufficient modeling of slippage is another problem; most existing solutions do not include slip ratio correction in velocity mapping or only perform control layer compensation, which may lead to insufficient accuracy during low-speed alignment. Furthermore, when dealing with a four-wheel platform including omnidirectional servo front wheels, existing ideal models struggle to correctly handle feasible attitude sets under nonholonomic constraints. Finally, most existing models and solutions lack online adaptive capabilities and a rapid response mechanism to sudden working conditions, such as sudden changes in load or road conditions, making it difficult to guarantee accuracy for certain critical tasks. Therefore, this invention aims to improve these shortcomings and provide a more precise and adaptable control method for an integrated intelligent wheelchair and bed-chair device. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology. To achieve the above objective, a wheelchair dynamics modeling method based on load sensing and slip compensation is adopted to solve the problems mentioned in the background technology.

[0006] A wheelchair dynamics modeling method based on load sensing and slip compensation includes the following steps: Acquire multi-source sensor data, which includes at least pressure data from a pressure array sensor arranged on the wheelchair seat, angular velocity data from a wheel speed encoder on the rear drive wheel, and data from an inertial measurement unit (IMU). Based on the seat pressure array data, the total mass and center of gravity of the wheelchair are determined, and the normal force of each wheel hub is calculated. An effective radius model and a stiffness model for the drive wheel hub are established. The effective radius model represents the functional relationship between the effective radius of the hub and the normal force, and the stiffness model represents the functional relationship between the longitudinal stiffness, lateral stiffness, and normal force of the hub. Based on the effective radius model of the drive wheel hub, angular velocity and linear velocity measurement data, a wheelchair kinematic model with slip compensation is established to calculate the corrected linear velocity and rotational angular velocity of the wheelchair. Construct a parameter vector to be identified that includes the effective radius model, stiffness model, and slip compensation-related parameters; The parameter vector to be identified is identified and updated online based on the recursive least squares method. The output of the kinematic model is corrected using the updated parameter vector; and The corrected model output is used for wheelchair pose estimation and motion control.

[0007] As a further aspect of the present invention: the determination of the total mass and center of gravity of the wheelchair based on seat pressure array data specifically includes: The total normal force exerted by the occupant on the seat cushion is calculated using the seat cushion pressure array sensor, and the occupant's mass is calculated by combining the gravitational acceleration value. Calculate the pressure center coordinates of the pressure array sensor; Based on the known coordinates of the wheelchair's center of mass and the coordinates of the pressure center, the offset of the center of mass in the vehicle coordinate system caused by the occupant's mass is calculated.

[0008] As a further aspect of the present invention: the kinematic model established by the fusion slip compensation specifically includes: The slip ratio of each drive wheel is calculated based on the effective radius of the hub, angular velocity, and linear velocity. The linear velocity and rotational angular velocity of the wheelchair calculated from the angular velocity are corrected using a slip correction function constructed based on the slip ratio.

[0009] As a further aspect of the present invention, the following steps are also included: A nonholonomic kinematic constraint is established for the omnidirectional motion of the front wheel, and the constraint is constructed with the condition that the lateral velocity of the front wheel is zero.

[0010] As a further aspect of the present invention: the parameter vector to be identified further includes the moment of inertia of the driving wheel and the axle damping coefficient; A dynamic model of the differential drive system is established based on the rotational dynamics equation of the drive wheel, and the parameters to be identified in the rotational dynamics equation are incorporated into the online identification and updating.

[0011] As a further aspect of the present invention: the online identification and updating of the parameter vector to be identified specifically includes: The initial estimate of the parameter vector to be identified is obtained through offline calibration experiments; During operation, the parameter vector is updated online using a recursive least squares method with a forgetting factor.

[0012] As a further aspect of the present invention: the online identification and update process also includes a residual triggering mechanism: Calculate the residual between the pose predicted by the model and the reference pose obtained from external observation; When the residual exceeds an adaptive threshold based on the standard deviation of historical residuals, high-gain identification is triggered to accelerate parameter convergence.

[0013] As a further aspect of the present invention: the step of using the corrected model output for wheelchair pose estimation specifically includes: The corrected wheelchair linear velocity and rotational angular velocity are input into an extended Kalman filter or an information filter; The data is fused with one or more data from an IMU, ultrasonic, or laser positioning sensor in the filter to output a high-precision fused pose estimate.

[0014] As a further aspect of the present invention: the step of using the modified model output for the motion control of the wheelchair specifically includes: In the predictive model of Model Predictive Control (MPC), the modified linear velocity and angular velocity are used as the system state update equations. The constraints of the MPC include a slip ratio limit, front wheel servo constraints, and environmental geometric collision constraints obtained based on the slip compensation model.

[0015] As a further aspect of the present invention: the influence of each parameter in the parameter vector to be identified on the system positioning error and slip triggering conditions is evaluated through offline calibration, multi-condition real vehicle tests and Monte Carlo simulation, and the forgetting factor, residual triggering threshold and initial value range of each parameter are set according to the evaluation results.

[0016] Compared with the prior art, the present invention has the following technical advantages: Using the aforementioned technical solution, the load distribution of the occupant and the motion state of the wheelchair are sensed in real time through multi-source sensors such as a seat pressure array, wheel speed encoder, and IMU. Based on this, a comprehensive kinematic model is established that simultaneously considers the changes in effective tire radius and stiffness caused by load, as well as the ground slip effect, and a parameter vector incorporating these relevant characteristics is constructed. Subsequently, a recursive least squares algorithm is used to continuously identify and update these model parameters online, thereby achieving adaptive correction of the model. Finally, the corrected model output is used to improve odometry accuracy and motion control performance. The beneficial effects of this method are that by integrating load sensing and slip compensation and introducing an online parameter self-learning mechanism, it effectively overcomes the model inaccuracy problems caused by changes in occupant weight, center of mass shift, and complex ground adhesion conditions, significantly improving the motion modeling accuracy, state estimation reliability, and trajectory tracking control performance of the wheelchair under different operating conditions. Attached Figure Description

[0017] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings: Figure 1 This is a schematic diagram illustrating the steps of the dynamic modeling method disclosed in this application. Figure 2 This is a flowchart of the dynamic modeling method disclosed in this application. Figure 3 This is a schematic diagram of the vehicle coordinate system according to an embodiment of this application. Detailed Implementation

[0018] 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.

[0019] Please refer to Figure 1 and Figure 2 In this embodiment of the invention, a wheelchair dynamics modeling method based on load sensing and slip compensation includes the following steps: Step S1: Acquire multi-source sensor data, which includes at least pressure data from pressure array sensors arranged on the wheelchair seat, angular velocity data from wheel speed encoders of the rear drive wheels, and data from an inertial measurement unit (IMU). In this embodiment, the total mass and center of gravity of the wheelchair are determined based on seat pressure array data, specifically including: The total normal force exerted by the occupant on the seat cushion is calculated using the seat cushion pressure array sensor, and the occupant's mass is calculated by combining the gravitational acceleration value. Calculate the pressure center coordinates of the pressure array sensor; Based on the known coordinates of the wheelchair's center of mass and the coordinates of the pressure center, the offset of the center of mass in the vehicle coordinate system caused by the occupant's mass is calculated.

[0020] In specific implementation methods, such as Figure 3 As shown, the diagram illustrates the vehicle's coordinate system; the wheelchair includes two rear differential drive wheels and two front omnidirectional follower wheels, in the global coordinate system. With vehicle coordinate system A wheelchair dynamics model is established below, where: the global coordinate system is used to describe the absolute pose of the wheelchair. ,origin The center of the headboard is the side of the bed, and the direction from the headboard to the footboard is... The axis, with the bed body facing the wheelchair. Axis; The vehicle coordinate system is used to describe wheelchair structural parameters, wheel speeds, center of mass, and mechanical quantities, with the origin at... Located in the center of the wheelchair, The axis points forward along the direction of the wheelchair's movement. The axis runs along the left side of the wheelchair.

[0021] Step S2: Based on the seat pressure array data, determine the total mass and center of gravity of the wheelchair, and calculate the normal force of each wheel hub; The steps for load sensing and hub normal force calculation are as follows: Calculate rider mass based on seat pressure array data and its centroid offset in the vehicle coordinate system And based on the known mass of the wheelchair body The distance from the front and rear wheels to the center of the wheelchair is Wheelbase Parameters, calculate the total mass and the normal force of the four wheels , i∈{left rear wheel RL, right rear wheel RR, left front wheel FL, right front wheel FR};

[0022]

[0023]

[0024]

[0025]

[0026] The wheelchair's structural dimensions and orientation are known. The IMU is a six-axis or nine-axis inertial measurement unit. The external positioning data is one or more combinations of ultrasonic positioning and laser positioning. By calculating the total normal force exerted by the occupant on the seat cushion using a seat pressure array sensor, and combining this with the gravitational acceleration value g, the occupant's mass can be calculated. Alternatively, the weight of the passenger can be measured by deploying weighing sensors. ; The seat pressure array sensor calculates its pressure center coordinates CoP( , );

[0027] in, The normal pressure at point i in row j of the pressure pad. , The coordinates of the pressure point in the wheelchair coordinate system.

[0028] Its characteristic is that it utilizes the pressure center CoP and the known wheelchair center of mass. Build passenger quality The resulting shift of the vehicle coordinate system's center of mass The relationship between the pressure center CoP;

[0029] Step S3: Establish the effective radius model and stiffness model of the drive wheel hub. The effective radius model represents the functional relationship between the effective radius of the hub and the normal force, and the stiffness model represents the functional relationship between the longitudinal stiffness, lateral stiffness and normal force of the hub. Among them, the effective radius and stiffness of the hub are modeled mechanically: An effective radius and stiffness model was established for the rear drive wheel of the wheelchair, and the effective radius of the wheel hub was constructed. With normal force The function;

[0030] in This is the reference radius of the wheel hub when the wheelchair is static and unloaded. The sensitivity coefficient of the effective radius of the wheel hub to changes in normal force. This is the rated normal reference load value for the wheel hub; Establish the longitudinal stiffness of the drive wheel hub Lateral stiffness With normal force The function expression is:

[0031] For the front swivel wheel, a simplified rolling damping model is adopted to construct the front wheel rolling damping force. The expression is:

[0032] in, , The linear fitting coefficient for the longitudinal stiffness of the wheel hub is denoted as . , The linear fitting coefficients for the lateral stiffness of the wheel hub are denoted as . The coefficient of rolling friction for the front wheel is denoted as .

[0033] Step S4: Based on the effective radius model of the drive wheel hub, angular velocity and linear velocity measurement data, establish a wheelchair kinematic model with fused slip compensation to calculate the corrected linear velocity and rotational angular velocity of the wheelchair. In this embodiment, the kinematic model is established by incorporating slip compensation, specifically including: The slip ratio of each drive wheel is calculated based on the effective radius and angular velocity of the wheel hub. The linear velocity and rotational angular velocity of the wheelchair calculated from the angular velocity are corrected using a slip correction function constructed based on the slip ratio.

[0034] Among them, the wheel hub kinematics modeling that incorporates slip compensation is as follows: Based on the angular velocity of the left and right rear drive wheels Calculate the slip ratio of the left and right rear drive wheels based on linear velocity. The expression is:

[0035] Introducing linear velocity slip correction factor Correction, calculation of the linear velocity of the moving wheelchair body. ;

[0036] Introducing angular velocity slip correction factor Correction, calculation of the rotational angular velocity of the movable wheelchair body. ;

[0037] Meanwhile, using the constraint that the lateral speed of the front wheel is zero as the constraint condition, a nonholonomic constraint expression for the front wheel following is constructed for model consistency. In the above equation, This represents the parallel linear velocity component at the wheel-ground contact point. To avoid using positive constants with a denominator of zero at low speeds, For the slip correction function, This is the average slip ratio of the left and right wheels. To account for the equivalent wheelbase between wheels after flexible deformation.

[0038] Slip correction function It is a saturated nonlinear function, preferably A linear approximation is used in the low slip range. .

[0039] The front wheels include the left front wheel and the right front wheel, located on either side of the front axle in the vehicle coordinate system. Their center point coordinates in the vehicle coordinate system are respectively... )and The longitudinal and lateral velocity components of the left and right front wheels are respectively and The front wheel follow-up constraint is the front wheel steering angle. Satisfying the following follower relationship equations

[0040] Step S5: Construct a parameter vector to be identified that includes the effective radius model, stiffness model, and slip compensation related parameters; The specific steps for constructing the dynamic model identification parameters are as follows: A dynamic model was established based on the Newton-Euler method and rolling constraints of vehicle dynamics, yielding the dynamic equations for the rotation of the left and right rear drive wheel hubs:

[0041]

[0042] in, For left and right wheel hubs Longitudinal force on the axis, It is a nonlinear friction torque function. For the motor output torque, This refers to the longitudinal force acting on the wheel hub; The total moment of inertia of the wheel hub. The wheel axle damping coefficient, wheel hub moment of inertia, and wheel axle damping coefficient are incorporated into the online parameter identification vector; Based on the hub effect radius model, stiffness model, slip compensation, and the aforementioned dynamic identification parameters, a model-identifiable parameter vector is constructed. Used for identification and parameter updates;

[0043] Among them, the initial estimate Obtained through offline calibration experiments. These are the moments of inertia of the left and right drive wheels, respectively. These are the damping coefficients for the left and right drive wheels, respectively.

[0044] Step S6: Perform online identification and updating of the parameter vector to be identified based on the recursive least squares method; In this embodiment, the online identification and updating of the parameter vector to be identified specifically includes: The initial estimate of the parameter vector to be identified is obtained through offline calibration experiments; During operation, the parameter vector is updated online using a recursive least squares method with a forgetting factor.

[0045] In this embodiment, the online identification and update process also includes a residual triggering mechanism: Calculate the residual between the pose predicted by the model and the reference pose obtained from external observation; When the residual exceeds an adaptive threshold based on the standard deviation of historical residuals, high-gain identification is triggered to accelerate parameter convergence.

[0046] In a specific implementation, the online identification steps for model parameters are as follows: During operation, the parameter vector is processed based on the recursive least squares (RLS) algorithm. The RLS is updated online and includes a forgetting factor. When the model predicts the pose and the residual of the externally observed pose... Exceeding the threshold High-gain identification is triggered at certain times to accelerate convergence, where The standard deviation of the residuals. This is a preset threshold.

[0047] Step S7: Correct the output of the kinematic model using the updated parameter vector; In a specific implementation, the model update and pose fusion steps are as follows: The updated model parameters and velocity mapping results are used to correct the odometry output, which is then fused with IMU, ultrasonic, or laser positioning sensor data in the global coordinate system to output the fused pose estimate.

[0048] Step S8: Output the corrected model for wheelchair pose estimation and motion control.

[0049] In a specific implementation, based on the modified kinematic model, an intelligent control algorithm is used to achieve high-precision navigation of the wheelchair and autonomous docking control of the bed-chair in a confined space.

[0050] This embodiment also includes the following steps: A nonholonomic kinematic constraint is established for the omnidirectional motion of the front wheel, and the constraint is constructed with the condition that the lateral velocity of the front wheel is zero.

[0051] In this embodiment, the corrected model output is used for wheelchair pose estimation, specifically including: The corrected wheelchair linear velocity With rotational angular velocity Input to an extended Kalman filter or information filter; The data is fused with one or more data from an IMU, ultrasonic, or laser positioning sensor in the filter to output a high-precision fused pose estimate, which is used for docking control closed-loop feedback.

[0052] In this embodiment, the corrected model output is used for the motion control of the wheelchair, specifically including: In the prediction model of Model Predictive Control (MPC), the corrected linear velocity is used. and angular velocity As the system state update equation; In the constraints of the MPC, slip ratio limits, front wheel servo constraints, and environmental geometric collision constraints obtained based on the slip compensation model are introduced to achieve high-precision trajectory tracking of the wheelchair.

[0053] In this embodiment, the influence of each parameter in the parameter vector to be identified on the system positioning error and slip triggering conditions is evaluated through offline calibration, multi-condition real vehicle test and Monte Carlo simulation. Based on the evaluation results, the RLS forgetting factor, residual triggering threshold and initial value range of each parameter are set.

[0054] Specifically, residuals trigger high-gain identification of residuals Defined in global coordinate system 、 Difference L2 Norm

[0055] in The wheelchair pose vector calculated for the model. Reference pose vector obtained by fusing external sensors.

[0056] The feature is that the threshold strategy triggered by residuals includes one based on the historical residual standard deviation. Adaptive threshold ,in The typical value is 3-5.

[0057] Initial parameter values ​​in the model are estimated using an offline calibration method. Specifically, through no-load, half-load, and full-load tests, as well as tests on different road conditions such as hard ground, wet and slippery surfaces, and low-friction pads, the wheel hub encoder, IMU, pressure array, and external positioning data can be analyzed, and initial parameter values ​​can be obtained through offline calibration, including the static radius. Initial value of radius sensitivity coefficient Tire stiffness ( Front wheel rolling damping Wheelbase Front and rear wheelbase ( ), wheel inertia ( ), hub damping ( Hub reference load value wait; The Recursive Least Squares (RLS) algorithm is used for online identification and updating. The forgetting factor used in the RLS The recommended value range is 0.98-0.995, in order to balance parameter tracking and steady-state noise suppression.

[0058] The beneficial effects of this invention are: Enhanced adaptability to different operating conditions: By estimating the center of mass and mass in real time through a pressure array and adjusting the normal force and tire parameters accordingly, the model can adapt to different passengers and sitting postures, significantly reducing errors caused by load changes.

[0059] More systematic slip compensation: The slip ratio is incorporated as part of the speed mapping correction, rather than being compensated locally at the control end, which improves the consistency between the odometer and the prediction model, and results in better low-speed fine-tuning and docking accuracy.

[0060] More comprehensive support for tasks in confined spaces: Front wheel servo constraints are introduced during modeling and used for trajectory feasibility assessment. Combined with parametric models, this provides accurate predictions for MPC and improves the success rate of docking in narrow spaces.

[0061] Faster online adaptive speed: Employing residual-triggered RLS high-gain mode, it can quickly re-estimate parameters under sudden working conditions such as abrupt load changes or road surface changes, and the time to restore model accuracy is shorter than that of traditional static models.

[0062] The overall system is highly robust: the model-recognition-fusion-control system forms a closed loop, which can collaboratively handle sensor noise, short-term slippage and slowly changing parameters, thereby improving docking stability and security.

[0063] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention. The scope of the invention is defined by the appended claims and their equivalents, all of which should be included within the scope of protection of the invention.

Claims

1. A wheelchair dynamics modeling method based on load sensing and slip compensation, characterized in that, Includes the following steps: Acquire multi-source sensor data, which includes at least pressure data from a pressure array sensor arranged on the wheelchair seat, angular velocity data from a wheel speed encoder on the rear drive wheel, and data from an inertial measurement unit (IMU). Based on the seat pressure array data, the total mass and center of gravity of the wheelchair are determined, and the normal force of each wheel hub is calculated. An effective radius model and a stiffness model for the drive wheel hub are established. The effective radius model represents the functional relationship between the effective radius of the hub and the normal force, and the stiffness model represents the functional relationship between the longitudinal stiffness, lateral stiffness, and normal force of the hub. Based on the effective radius model of the drive wheel hub, angular velocity and linear velocity measurement data, a wheelchair kinematic model with slip compensation is established to calculate the corrected linear velocity and rotational angular velocity of the wheelchair. Construct a parameter vector to be identified that includes the effective radius model, stiffness model, and slip compensation-related parameters; The parameter vector to be identified is identified and updated online based on the recursive least squares method. The output of the kinematic model is corrected using the updated parameter vector; as well as The corrected model output is used for wheelchair pose estimation and motion control.

2. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 1, characterized in that, The determination of the total mass and center of gravity of the wheelchair based on seat pressure array data specifically includes: The total normal force exerted by the occupant on the seat cushion is calculated using the seat cushion pressure array sensor, and the occupant's mass is calculated by combining the gravitational acceleration value. Calculate the pressure center coordinates of the pressure array sensor; Based on the known coordinates of the wheelchair's center of mass and the coordinates of the pressure center, the offset of the center of mass in the vehicle coordinate system caused by the occupant's mass is calculated.

3. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 1, characterized in that, The kinematic model established by the fusion slip compensation specifically includes: Calculate the slip ratio of each drive wheel based on the effective radius, angular velocity, and linear velocity of the wheel hub; The linear velocity and rotational angular velocity of the wheelchair calculated from the angular velocity are corrected using a slip correction function constructed based on the slip ratio.

4. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 1, characterized in that, It also includes the following steps: A nonholonomic kinematic constraint is established for the omnidirectional motion of the front wheel, and the constraint is constructed with the condition that the lateral velocity of the front wheel is zero.

5. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 1, characterized in that, The parameter vector to be identified also includes the moment of inertia of the driving wheel and the axle damping coefficient; A dynamic model of the differential drive system is established based on the rotational dynamics equation of the drive wheel, and the parameters to be identified in the rotational dynamics equation are incorporated into the online identification and updating process.

6. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 5, characterized in that, The online identification and updating of the parameter vector to be identified specifically includes: The initial estimate of the parameter vector to be identified is obtained through offline calibration experiments; During operation, the parameter vector is updated online using a recursive least squares method with a forgetting factor.

7. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 6, characterized in that, The online identification and update process also includes a residual triggering mechanism: Calculate the residual between the pose predicted by the model and the reference pose obtained from external observation; When the residual exceeds an adaptive threshold based on the standard deviation of historical residuals, high-gain identification is triggered to accelerate parameter convergence.

8. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 1, characterized in that, The step of using the corrected model output for wheelchair pose estimation specifically includes: The corrected wheelchair linear velocity and rotational angular velocity are input into an extended Kalman filter or an information filter; The data is fused with one or more data from an IMU, ultrasonic, or laser positioning sensor in the filter to output a high-precision fused pose estimate.

9. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 8, characterized in that, The step of using the corrected model output for wheelchair motion control specifically includes: In the predictive model of Model Predictive Control (MPC), the modified linear velocity and angular velocity are used as the system state update equations. The constraints of the MPC include a slip ratio limit, front wheel servo constraints, and environmental geometric collision constraints obtained based on the slip compensation model.

10. The wheelchair dynamics modeling method based on load sensing and slip compensation according to claim 1, characterized in that, Through offline calibration, multi-condition real vehicle tests and Monte Carlo simulation, the influence of each parameter in the parameter vector to be identified on the system positioning error and slip triggering conditions is evaluated. Based on the evaluation results, the forgetting factor, residual triggering threshold and initial value range of each parameter for the recursive least squares method are set.

Images