Flexible surface-based magnetic particle array XR motion platform control method and system

Abstract

The application provides a kind of based on flexible surface multifunctional magnetic micro-particle array XR movement platform control method and system, method includes: collecting magnetic micro-particle unit pressure, two feet inertia measurement data, optical motion capture posture and sole magnetic force feedback data, after time synchronization, coordinate conversion, abnormality check and credibility empowerment generate unified perception data flow;Based on pressure data to establish continuous pressure field, analyze foot pressure gravity center, two feet pressure ratio, foot contact area, peak pressure and gait state;User horizontal position is estimated with extended Kalman filter, user speed estimation vector, user step speed scalar, user motion direction estimation value and user displacement trend vector;Read local terrain height map and material label map in XR scene, generate target terrain height field, smooth target height set and target resistance parameter vector;Construct the joint optimization problem of terrain simulation, resistance feedback and displacement compensation.

Description

Technical Field

[0001] This invention relates to the field of extended reality (XR) human-computer interaction and motion simulation control technology, and in particular to a control method and system for an XR motion platform based on a magnetic particle array with a flexible surface. Background Technology

[0002] Extended reality (XR) motion platforms are used to provide users with interactive motion experiences such as walking, turning, going uphill, going downhill, treading ice, and treading sand in virtual scenes. Existing XR motion platforms mostly employ omnidirectional treadmills, restraint devices, rigid lifting platforms, or combinations of multiple rigid load-bearing units, achieving motion feedback through mechanical sliding, lifting, vibration, or preset scene commands. This type of solution can meet basic needs for walking in place or simple motion simulation, but the platform surface is usually a rigid structure, making it difficult to continuously reconstruct flexible terrain and to create fine-grained, programmable foot resistance feedback for different ground materials.

[0003] Existing control methods suffer from limitations such as limited perception dimensions, fragmented control chains, and insufficient real-time performance. On one hand, some systems rely primarily on inertial measurement units (IMUs), optical motion capture data, or scene scripts to drive platform actions, lacking unified integration of magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture posture, and sole magnetic feedback data. This results in insufficient accuracy in recognizing user movement intentions and poor stability under conditions of motion capture occlusion, inertial drift, or rapid changes in sole contact. On the other hand, terrain simulation, drag feedback, and displacement compensation are typically controlled independently by different modules, lacking unified constraints and joint optimization among control objectives. This can easily lead to issues such as variations in support height, inconsistent magnetic drag output, and misalignment of displacement compensation directions.

[0004] Furthermore, existing XR motion platforms do not adequately utilize the semantic and material differences of virtual terrain, often struggling to automatically generate platform target shapes and resistance parameters based on local terrain height maps, material label maps, and the user's real-time gait. When the virtual scene switches between ice, sand, mud, rocks, steps, or slopes, platform feedback is prone to delays, abrupt changes, or inconsistencies with the user's foot perception, reducing immersion and safety. Therefore, this invention proposes a control method and system for an XR motion platform based on a magnetic particle array with a flexible surface.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a control method and system for a magnetic microparticle array XR motion platform based on a flexible surface, thereby solving the technical problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A control method for an XR motion platform based on a magnetic microparticle array using flexible surfaces includes the following steps: S1. Collect magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture posture and shoe sole magnetic feedback data, and perform time synchronization, coordinate transformation, anomaly verification and credibility weighting on each data source to generate a unified perception data stream; S2. Establish a continuous pressure field based on the pressure data in the unified sensing data stream, calculate the plantar pressure center of gravity, bipedal pressure ratio, plantar contact area and peak pressure, and analyze to obtain gait state and gait feature vector; S3. Based on the unified perception data stream and gait feature vector, the extended Kalman filter is used to estimate the user's horizontal position, user velocity estimation vector, user gait scalar, user movement direction estimation value, and user displacement trend vector to generate motion intention parameters. S4. Read the local terrain height map and material label map in the XR scene according to the motion intention parameters, complete the mapping between virtual coordinates and platform center coordinates, and generate the target terrain height field, smooth target height set and target resistance parameter vector. S5. Based on the motion intention parameters and target shape data, construct a joint optimization problem of terrain simulation, resistance feedback and displacement compensation, and solve the height command of the liftable support unit and the current and magnetic pole direction control vector of the magnetic particle unit under the constraints of stroke, current, safety force and rate of change. S6. Based on the real-time operating system, the platform executes instructions in a hierarchical closed-loop scheduling manner, performs magnetic field updates, supports altitude control, safety monitoring, anomaly degradation and emergency stop protection, and outputs execution feedback data; S7. Update the personalized parameter set based on gait feature vectors, motion intention parameters, target morphology data, execution feedback data, and user feedback scores, and feed the personalized parameter set back to the target morphology generation and joint optimization control process.

[0008] S1 specifically includes: establishing a platform center coordinate system with the geometric center of the platform's flexible surface; collecting magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture posture, and shoe sole magnetic feedback data; aligning the data from each path to a unified time axis based on hardware timestamps; and transforming the data from each path to the platform center coordinate system through rotation matrices and translation vectors; verifying timestamps, pressure jumps, inertial measurement anomalies, motion capture occlusion, and magnetic feedback anomalies; calculating sensor reliability weights based on anomaly errors; and generating a unified sensing data stream.

[0009] S2 specifically includes: establishing a continuous pressure field by bilinear interpolation based on the pressure output value and geometric position of the magnetic microparticle unit; calculating the plantar pressure center of gravity, the total pressure of the left foot, the total pressure of the right foot, the pressure ratio of both feet, the plantar contact area, and the peak pressure based on the continuous pressure field; identifying the right foot support state, the left foot support state, the double foot support state, the foot lift state, and the foot landing state based on the pressure ratio of both feet, the plantar contact area, the peak pressure, and the pressure change rate, and generating a gait feature vector.

[0010] S3 specifically includes: constructing an extended Kalman filter state vector using the user's horizontal position, user's horizontal velocity vector, user's forward direction angle, and inertial measurement unit acceleration bias; forming a joint observation vector by combining pressure center of gravity displacement observation, optical motion capture hip joint position observation, and inertial measurement short-time integral displacement observation; adjusting the observation noise covariance and process noise covariance according to sensor confidence weights and gait state, and outputting the user velocity estimation vector, user gait scalar, user motion direction estimation value, and user displacement trend vector to form motion intention parameters.

[0011] S4 specifically includes: reading local terrain height maps, material label maps, and scene semantic labels through the XR scene interface; establishing a mapping relationship between XR virtual coordinates and platform center coordinates based on the virtual-to-real ratio coefficient and rotation alignment matrix; extracting the local virtual terrain corresponding to the current foot contact area and the predicted foot landing area based on the user's horizontal position, estimated user movement direction, and user scalar step speed, and generating smooth target heights for each liftable support unit; and querying the material and resistance mapping table based on the material label of the foot contact area to generate a target resistance parameter vector.

[0012] S5 specifically includes: constructing the actual height field of the platform based on the target terrain height field and the height command of the liftable support unit; calculating the actual magnetic output based on the current command of the magnetic particle unit, the magnetic pole direction control vector, the magnetic gain coefficient, and the location of the foot action area; calculating the desired magnetic feedback based on the target resistance parameter vector, the user's normal foot force, and the user's velocity estimation vector; constructing a joint optimization objective based on terrain simulation error, resistance feedback deviation, displacement compensation error, height change smoothing term, and current change smoothing term, and solving the platform execution command under the constraints of travel, current, safety force, height change rate, and current change rate.

[0013] S6 specifically includes: a 1000Hz real-time layer, a 500Hz control layer, a 100Hz strategy layer, and a 10Hz learning layer based on a real-time operating system; the 1000Hz real-time layer performs magnetic field command updates, safety monitoring, and emergency stop protection; the 500Hz control layer performs pressure field sampling, gait state updates, and extended Kalman filter state updates; the 100Hz strategy layer performs virtual terrain parsing, joint optimization solutions, and XR scene engine data synchronization; and the 10Hz learning layer performs personalized parameter evaluation. When current over-limit, magnetic anomaly, fall risk, or control link anomaly occurs, magnetic field attenuation, support height safety retraction, and XR scene pause are executed.

[0014] S7 specifically includes: constructing an online learning state vector from gait feature vectors, motion intention parameters, target morphology data, execution feedback data, and user feedback scores; building a reward function from terrain simulation error, resistance feedback error, displacement compensation error, gait naturalness score, and user comfort score; outputting the terrain simulation error weight, resistance feedback deviation weight, displacement compensation error weight, displacement compensation coefficient, and parameter adjustment amounts for the material and resistance mapping table; and forming a personalized parameter set after performing safety boundary checks and degradation rollback on candidate parameters and feeding it back to the target morphology generation and joint optimization control process.

[0015] The control system for the XR motion platform based on a flexible surface magnetic microparticle array includes: a multi-source heterogeneous sensing data acquisition and spatiotemporal synchronization module, a pressure field modeling and gait analysis module, a motion intent estimation module, a virtual terrain analysis and target shape generation module, a joint optimization control module, a high-frequency closed-loop scheduling and safety protection module, and an online learning and personalized adaptive optimization module.

[0016] The beneficial effects of this invention are as follows: This invention integrates magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture posture, and sole magnetic feedback data to maintain the stability of motion intention estimation when motion capture is obstructed, inertial drift occurs, or the sole contact state changes, thereby improving the recognition accuracy of XR motion platforms for user walking, turning, and foot landing actions.

[0017] This invention analyzes the plantar pressure center of gravity, bipedal pressure ratio, plantar contact area, and gait state based on a continuous pressure field, enabling platform control to generate feedback that follows the user's actual plantar force changes. This reduces feedback lag and misjudgments caused by relying solely on scene scripts or single sensors. By generating a smooth target height set and target resistance parameter vector based on local terrain height maps and material label maps, it can provide differentiated terrain support and plantar resistance feedback in virtual scenes such as ice, sand, mud, rocks, steps, and slopes, improving the consistency of the virtual terrain experience.

[0018] This invention unifies terrain simulation, resistance feedback, and displacement compensation into a joint optimization problem with safety constraints. This coordinates the height command of the liftable support unit with the current-pole direction control vector of the magnetic particle unit, avoiding command conflicts caused by the independent execution of multiple control objectives. Through hierarchical closed-loop scheduling, safety monitoring, anomaly degradation, and personalized parameter updates, the XR motion platform meets real-time control requirements while continuously correcting control parameters based on user gait habits and feedback scores, improving long-term safety, comfort, and adaptability. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the XR motion platform control method based on a flexible surface magnetic microparticle array according to the present invention. Detailed Implementation

[0020] 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.

[0021] Example 1: As Figure 1 As shown, this embodiment provides a control method for an XR motion platform based on a flexible surface magnetic microparticle array, including the following steps: S1. Collect magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture posture and shoe sole magnetic feedback data, and perform time synchronization, coordinate transformation, anomaly verification and credibility weighting on each data source to generate a unified perception data stream; S2. Establish a continuous pressure field based on the pressure data in the unified sensing data stream, calculate the plantar pressure center of gravity, bipedal pressure ratio, plantar contact area and peak pressure, and analyze to obtain gait state and gait feature vector; S3. Based on the unified perception data stream and gait feature vector, the extended Kalman filter is used to estimate the user's horizontal position, user velocity estimation vector, user gait scalar, user movement direction estimation value, and user displacement trend vector to generate motion intention parameters. S4. Read the local terrain height map and material label map in the XR scene according to the motion intention parameters, complete the mapping between virtual coordinates and platform center coordinates, and generate the target terrain height field, smooth target height set and target resistance parameter vector. S5. Based on the motion intention parameters and target shape data, construct a joint optimization problem of terrain simulation, resistance feedback and displacement compensation, and solve the height command of the liftable support unit and the current-magnetic pole direction control vector of the magnetic particle unit under the constraints of stroke, current, safety force and rate of change. S6. Based on the real-time operating system, the platform executes instructions in a hierarchical closed-loop scheduling manner, performs magnetic field updates, supports altitude control, safety monitoring, anomaly degradation and emergency stop protection, and outputs execution feedback data; S7. Update the personalized parameter set based on gait feature vectors, motion intention parameters, target morphology data, execution feedback data, and user feedback scores, and feed the personalized parameter set back to the target morphology generation and joint optimization control process.

[0022] Specifically as follows: S1. After the XR motion platform is started, the control system performs sensor calibration, platform coordinate calibration, and user initial standing calibration to establish the platform's central coordinate system. The platform's central coordinate system uses the geometric center of the flexible surface as the origin, the longitudinal direction of the platform corresponding to the user's initial forward direction as the ordinate axis, the transverse direction of the platform perpendicular to the user's initial forward direction as the abscissa axis, and the platform normal as the height direction. Subsequent calculations based on magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture attitude, shoe sole magnetic feedback, virtual terrain mapping data, and the height of the adjustable support unit are all performed using the platform's central coordinate system.

[0023] The control system reads the pressure output value of each magnetic particle unit in the magnetic particle array at a sampling frequency of 1000Hz to form the original pressure data vector: ; in, This represents the original pressure data vector at time t. This represents the pressure output value of the i-th magnetic particle unit at time t. This indicates the total number of magnetic particle units.

[0024] The control system reads the acceleration and angular velocity vectors output by the bipedal inertial measurement unit at a sampling frequency of 200Hz, the whole-body joint posture quaternions output by the optical motion capture system at a sampling frequency of 120Hz, and the magnetic feedback vector output by the magnetic feedback sensing layer of the sole at a sampling frequency of 500Hz. All of the above data are accompanied by hardware timestamps and use a local high-precision clock as a unified time reference.

[0025] For data with different sampling frequencies, the control system aligns the sensor data from each source to a unified time axis. Low-frequency data is compensated using linear interpolation to the current control cycle, while high-frequency data is selected from the data closest to the target time within the corresponding time window, or the window average data is used. The time synchronization error of multi-source data satisfies: ; in, This represents the time synchronization error between multi-source sensor data.

[0026] When any sensor data has missing timestamps, timestamps that are backwards, sampling intervals that exceed twice the corresponding theoretical sampling period, or data arrival times that exceed the current control period, the control system marks that data as time-abnormal data and prohibits it from directly entering the fusion calculation of the current period.

[0027] The control system converts the output data of the SSS-type sensor to the platform center coordinate system based on the pre-calibration results. ; in, This represents the position vector transformed to the platform's central coordinate system. This represents the position vector output by the s-th type of sensor in its own coordinate system. This represents the rotation matrix from the coordinate system of the s-th type of sensor to the coordinate system of the platform center. Let represent the translation vector from the coordinate system of the s-th type of sensor to the coordinate system of the platform center, where s represents the sensor type.

[0028] After completing time synchronization and coordinate transformation, the control system generates a unified sensing data stream: ; in, Represents the unified sensing data stream at time t. This represents the acceleration vector output by the bipedal inertial measurement unit and transformed into the platform's central coordinate system. This represents the angular velocity vector output by the bipedal inertial measurement unit and converted to the platform's central coordinate system. This represents the quaternion representing the whole-body joint posture output by the optical motion capture system. This represents the magnetic feedback vector output by the magnetic feedback sensing layer of the shoe sole. Indicates a unified timestamp. Let represent the confidence weight of the s-th type of sensor at time t.

[0029] The control system verifies the validity of each data stream in the unified sensing data stream. For pressure data, if the pressure value of a single magnetic particle unit changes by more than three times the standard deviation relative to the average of the previous five frames, or if the pressure value exceeds the upper limit of the pressure sensor's range, the corresponding pressure point is marked as a pressure anomaly. For inertial measurement data, if the acceleration modulus exceeds the upper limit of human motion capability determined by the user's initial calibration or system safety parameters, or if the angular velocity modulus exceeds the preset upper limit of the steering angular velocity, the frame of inertial measurement data is marked as inertial anomaly data.

[0030] For optical motion capture data, when the same key joint marker is lost for more than 3 consecutive frames, or when the displacement of the joint marker in adjacent frames exceeds the upper limit of human movement ability, the corresponding motion capture data will be marked as abnormal motion capture data. For magnetic feedback data, when the feedback channel is disconnected, the feedback value remains 0, or the feedback value exceeds the user's safety force threshold, the corresponding magnetic feedback data will be marked as abnormal magnetic data.

[0031] The control system calculates sensor reliability weights based on abnormal errors: ; in, This represents the confidence weight of the s-th type of sensor at time t. This indicates the current abnormal error of the s-th type of sensor. This indicates the maximum permissible abnormal error for the s-th type of sensor.

[0032] when When the data level falls below a preset confidence threshold, the control system reduces the weight of that type of sensor in subsequent filtered observations. When the number of consecutive preset periods for that type of sensor falls below the confidence threshold, the control system temporarily removes the corresponding observation dimension from the fusion model and activates available redundant sensing channels. The unified sensing data stream output by S1... Used for S2 pressure field modeling, S3 motion intent estimation, and S6 safety monitoring.

[0033] S2, The control system receives the unified sensing data stream output by S1. Read the raw pressure data vector from it. A continuous two-dimensional pressure field is established based on the geometric position of the magnetic particle unit in the coordinate system at the center of the platform.

[0034] For the i-th magnetic particle unit, its platform coordinates are: The control system uses bilinear interpolation to convert discrete pressure sampling points into a continuous pressure field. ; in, This represents the coordinates in the platform's central coordinate system at time t. Continuous pressure value at the location, This represents the pressure output value of the i-th magnetic particle unit at time t. Indicates the position of the i-th magnetic particle unit Bilinear basis functions centered at , This indicates the total number of magnetic particle units.

[0035] Bilinear basis functions satisfy: ; in, Let i represent the i-th bilinear basis function.

[0036] To distinguish between the left and right foot contact areas, the control system records the initial contact areas of the left and right feet during the user's initial standing calibration phase. It then uses the lateral coordinates of the pressure center of gravity, the foot contact profile, and shoe sole identification marks to divide the pressure field into regions. Magnetic microparticles located in the left foot contact area with pressure values ​​exceeding the contact threshold participate in the left foot pressure calculation, while those located in the right foot contact area with pressure values ​​exceeding the contact threshold participate in the right foot pressure calculation.

[0037] The control system calculates the position of the plantar pressure center of gravity: ; in, This indicates the position of the center of gravity of the plantar pressure at time t. This represents the lateral coordinate of the i-th magnetic particle unit in the coordinate system at the center of the platform. This represents the longitudinal coordinate of the i-th magnetic particle unit in the coordinate system at the center of the platform. This indicates the prevention of extremely small positive numbers with a denominator of 0.

[0038] When the total pressure in the current cycle is lower than the total contact threshold, the control system uses the previous effective pressure center position and marks the current cycle as a control cycle with low contact reliability.

[0039] The control system calculates the sum of the pressure on the left foot and the sum of the pressure on the right foot, and obtains the pressure ratio between the two feet: ; in, This represents the ratio of pressure on both feet at time t. This represents the total pressure on the left foot. This represents the total pressure on the right foot. This indicates the prevention of extremely small positive numbers with a denominator of 0.

[0040] The control system calculates the plantar contact area: ; in, This represents the area of ​​foot contact at time t. This indicates the number of magnetic particle units whose pressure value exceeds the contact threshold. This represents the effective area corresponding to a single magnetic particle unit.

[0041] The control system calculates the peak pressure: ; in, This represents the peak pressure at time t. This represents the pressure output value of the i-th magnetic particle unit. This indicates the total number of magnetic particle units.

[0042] The control system constructs a gait state machine based on the position of the center of gravity of plantar pressure, the ratio of pressure between the two feet, the plantar contact area, peak pressure, and the rate of pressure change. The gait state is defined as follows: ; in, This represents the gait state at time t.

[0043] when When the total pressure on the left foot is higher than the contact threshold and the total pressure on the right foot is lower than the contact threshold, the control system determines that the left foot is in a supporting state. When the total pressure on the right foot is higher than the contact threshold and the total pressure on the left foot is lower than the contact threshold, the control system determines that the right foot is in a supporting state. When the total pressure of the left foot and the total pressure of the right foot are both higher than the contact threshold, the control system determines that the system is in a bipedal support state.

[0044] When the pressure on any foot changes from above the contact threshold to below the contact threshold, and the contact area on the sole of the foot continuously decreases, the control system determines that the foot is in a lifting state. When the pressure on any foot changes from below the contact threshold to above the contact threshold, and the peak pressure rise rate exceeds the landing threshold, the control system determines that the foot is in a landing state.

[0045] To accommodate the gait differences among users, the control system employs a dynamic time warping algorithm to match the current pressure center of gravity trajectory with a gait template. The gait template is generated from system-preset standard gait samples, user-initialized gait samples, or historical stable gait data fragments. When the dynamic time warping matching result matches the bipedal pressure ratio threshold judgment result, the control system directly outputs the corresponding gait state; when the two results are inconsistent, the control system determines the final gait state based on the stability of the pressure data and the sensor reliability weights output by S1.

[0046] Output gait feature vector: ; in, This represents the gait feature vector at time t. and Indicates the coordinates of the center of gravity of the plantar pressure. Indicates the bipedal pressure ratio. Indicates gait state, Indicates the contact area of ​​the sole of the foot. This represents peak pressure. Gait feature vectors are used for S3 motion intention estimation, S4 foot contact area determination, S5 resistance feedback target calculation, and S7 gait naturalness evaluation.

[0047] S3, The control system receives the unified sensing data stream output by S1. and the gait feature vector output by S2 An extended Kalman filter is used to estimate the user's motion intention. The motion intention includes the user's horizontal position, the estimated user velocity vector, the user's gait scalar, the estimated user's direction of motion, the user's displacement trend vector, and the current gait state.

[0048] The state vector of the extended Kalman filter is defined as follows: ; in, This represents the state vector for the k-th filtering cycle. This indicates the user's horizontal position in the platform's central coordinate system. Represents the user's horizontal velocity vector. Indicates the user's forward direction angle. This indicates the acceleration bias of the inertial measurement unit.

[0049] The filter period is The bias-free acceleration input is defined as: ; in, This represents the debiasing acceleration input in the k-th filtering cycle. This represents the horizontal acceleration vector measured by the inertial measurement unit and converted to the platform center coordinate system during the k-th filtering cycle. This represents the acceleration bias in the k-th filter cycle.

[0050] The state prediction relationship is as follows: ; in, This indicates the user's horizontal position in the (k+1)th filtering cycle. This indicates the user's horizontal position in the k-th filtering cycle. This represents the user's horizontal velocity vector in the k-th filtering cycle. This indicates the biased acceleration input. Indicates the filter period.

[0051] ; in, This represents the user's horizontal velocity vector in the (k+1)th filtering cycle. This represents the user's horizontal velocity vector in the k-th filtering cycle. This indicates the biased acceleration input. Indicates the filtering period.

[0052] ; in, This represents the user's forward direction angle in the (k+1)th filtering cycle. This represents the user's forward direction angle in the k-th filtering cycle. This represents the rate of change of the direction angle obtained from the angular velocity of the inertial measurement unit or the attitude change of the optical motion capture system. Indicates the filtering period.

[0053] ; in, This represents the acceleration bias in the (k+1)th filter cycle. This represents the acceleration bias in the k-th filter cycle.

[0054] The above state prediction relationship can be written as: ; in, This represents the state vector for the (k+1)th filtering cycle. This represents the state transition matrix for the k-th filtering cycle. This represents the input matrix for the k-th filtering period. This indicates the biased acceleration input. This indicates process noise.

[0055] Process noise satisfies: ; in, Indicates process noise. Let represent the process noise covariance matrix for the k-th filtering cycle.

[0056] The observation vector consists of pressure center of gravity displacement observations, optical motion capture hip joint position observations, and inertial measurement short-time integral displacement observations. The pressure center of gravity displacement observations are as follows: ; in, This represents the pressure center of gravity displacement observation during the k-th filtering cycle. and These represent the coordinates of the plantar pressure centroid during the k-th filtering cycle.

[0057] Optical motion capture of hip joint position observation: ; in, This represents the optical motion capture hip joint position observation during the k-th filter cycle. and This represents the horizontal coordinate of the hip joint after its position has been transformed to the platform's central coordinate system.

[0058] Inertial measurement short-time integral displacement observation is as follows: ; in, This represents the short-time integral displacement observation of the inertial measurement during the k-th filter cycle. and This represents the horizontal displacement coordinate obtained by short-time integration using an inertial measurement unit.

[0059] The joint observation vector is: ; in, This represents the joint observation vector for the k-th filtering period. This indicates the observation of the displacement of the center of gravity under pressure. This indicates optical motion capture of hip joint position observation. This indicates short-time integral displacement observations from inertial measurements.

[0060] The control system dynamically configures the observation matrix based on the sensor confidence weights output by S1. When the reliability of the pressure data is lower than a preset threshold, the control system increases the noise in the pressure centroid observation; when optical motion capture data is obstructed or experiences continuous frame loss, the control system adjusts the observation matrix. The corresponding rows of optical motion capture observations are removed; when the drift of the inertial measurement unit exceeds the preset threshold, the control system increases the noise of the short-time integral displacement observation of the inertial measurement.

[0061] The observation noise covariance is updated according to the following formula: ; in, This represents the observation noise covariance of the s-th type sensor in the k-th filtering period. This represents the initial observation noise covariance of the s-th type of sensor. This represents the confidence weight of the s-th type of sensor at time t. This indicates the prevention of extremely small positive numbers with a denominator of 0.

[0062] Process noise covariance matrix Based on gait status Adaptive adjustment. When in a bipedal support state, the control system reduces position prediction noise and velocity prediction noise; when in a raised or lowered state, the control system increases velocity prediction noise; during control cycles with low contact reliability, the control system increases pressure center of gravity observation noise and increases the weight of short-time inertial measurement observations.

[0063] After completing the extended Kalman filter prediction and update, the control system outputs the user velocity estimation vector: ; in, This represents the user velocity estimation vector at time t. This represents the user's horizontal velocity vector during the k-th filtering cycle.

[0064] The control system outputs the user's step speed scalar: ; in, The scalar representing the user's step speed at time t. Represents the user velocity estimation vector. Represents the vector norm.

[0065] The control system outputs motion intention parameters: ; in, The parameter representing the motion intention at time t, Indicates the user's horizontal position. Represents the user velocity estimation vector. This represents a scalar value indicating the user's step speed. This represents the estimated direction of user movement. Represents the user's displacement trend vector. This indicates gait status. Motion intent parameters are used for S4 virtual terrain local area capture, S5 displacement compensation and resistance direction calculation, S6 fall risk assessment, and S7 personalized learning evaluation.

[0066] S4, The control system receives the motion intention parameters output by S3. The system reads local virtual environment data near the user's current location from the XR scene engine via a standardized XR scene interface. This standardized XR scene interface is implemented using REST API, shared memory, or a low-latency message queue, with a read frequency of 100Hz. The read data includes the user's virtual coordinates, local terrain height map, material label map, scene semantic labels, and terrain update version number.

[0067] The local topographic elevation map is represented as follows: ; in, Represents virtual coordinates in an XR virtual scene The terrain elevation is indicated by u and v, which represent the horizontal coordinates in the XR virtual scene.

[0068] Material label images are represented as follows: ; in, Represents virtual coordinates in an XR virtual scene Material labels are provided for various terrains, including grass, mud, sand, ice, rocks, planks, steps, and slopes.

[0069] The control system establishes a mapping relationship between XR virtual coordinates and platform center coordinates: ; in, This represents the platform coordinates in the platform's central coordinate system. Represents the coordinates of the XR virtual scene. Indicates a virtual coordinate reference point. Indicates the platform coordinate reference point. Indicates the ratio of virtual to real data. This represents the rotation alignment matrix between the virtual direction and the platform direction.

[0070] The virtual-to-real ratio coefficient and rotation alignment matrix are calibrated during system initialization and corrected based on the user's height, stride, and effective platform area. The control system adjusts the parameters based on the user's horizontal position. User movement direction estimation value and user step speed scalar It extracts a local virtual terrain region centered on the current foot contact area and the predicted foot landing area. The predicted foot landing area is determined based on the current gait state, the user's velocity estimation vector, and the user's movement direction estimation value, and is used to generate the target height of the liftable support unit before the user lands.

[0071] For the j-th liftable support unit, its platform coordinates are: The control system determines the corresponding virtual coordinates based on the mapping relationship between virtual and real coordinates. The target height is obtained by bilinear interpolation on the local terrain height map. ; in, This represents the target height of the j-th liftable support unit at time ttt. Represents the bilinear interpolation function. This represents a local terrain elevation map. This represents the virtual coordinates corresponding to the j-th liftable support unit. This indicates the amount of compensation for the platform's reference height offset.

[0072] To avoid foot impact caused by sudden height changes, the control system smooths out the target height: ; in, This represents the smoothed target height of the j-th liftable support unit at time t. This represents the target height of the j-th liftable support unit at time t. This indicates the smoothed target height of the previous control cycle. Indicates the height update smoothing coefficient, Indicates the control cycle.

[0073] When the terrain features significant height variations, such as steps, rocks, or slopes, the control system employs different height update rates for contact and non-contact areas based on gait status. The height response speed is increased for the predicted landing area and decreased for non-contact areas to balance terrain realism and foot safety.

[0074] The control system queries the material-resistance mapping table based on the material label corresponding to the sole contact area to generate a target resistance parameter vector. ; in, This represents the target resistance parameter vector at time t. This represents a material-resistance mapping table. This indicates the material label corresponding to the virtual coordinates of the plantar pressure center. This represents the virtual coordinates of the plantar pressure center of gravity after it has been mapped onto the XR virtual scene.

[0075] The target drag parameter vector includes the static friction coefficient, dynamic friction coefficient, normal drag gain, lateral drag gain, and hysteresis drag coefficient. Ice corresponds to a lower static friction coefficient and a lower dynamic friction coefficient; sand corresponds to a higher normal drag gain and a higher lateral drag gain; mud corresponds to a higher hysteresis drag coefficient; rocks correspond to local height disturbances and non-uniform drag gains; and wooden planks correspond to a medium friction coefficient and a lower hysteresis drag coefficient. These parameters are initialized by system preset values ​​and can be corrected through S7 online learning and personalized adaptive optimization.

[0076] When the sole contact area is covered by multiple material labels, the control system weights and fuses the target resistance parameter vector according to the area ratio of different materials in the sole contact area. S4 outputs target morphology data: ; in, Represents the target shape data at time t. This represents the set of smooth target heights for all liftable support units. Represents the target terrain elevation field. Represents the target resistance parameter vector. This represents the set of material labels for the foot contact area. Target morphology data serves as the input for S5 joint optimization control.

[0077] S5, The control system receives the motion intention parameters output by S3. and the target shape data output by S4 The terrain simulation, drag feedback, and displacement compensation are constructed as a constrained joint optimization problem, and the platform execution commands are solved within each policy control cycle. The platform execution commands include the height command of the liftable support unit and the current-pole direction control vector of the magnetic particle unit.

[0078] The height command vector for the liftable support unit is defined as follows: ; Where h represents the height command vector of the liftable support unit. This indicates the height command for the j-th liftable support unit. This indicates the total number of liftable support units.

[0079] The actual height of the platform is obtained by interpolating the height of the liftable support unit: ; in, This represents the actual height field of the platform at time t. denoted by , represents the interpolation mapping matrix from the height of the liftable support unit to the height field of the flexible surface, and h represents the height command vector of the liftable support unit.

[0080] The current-pole direction control vector of a magnetic particle unit is defined as: ; in, This represents the current-pole direction control vector of the magnetic particle unit. This represents the current command for the i-th magnetic particle unit. Let represent the magnetic pole direction control vector of the i-th magnetic particle unit. This indicates the total number of magnetic particle units.

[0081] The actual magnetic force output is calculated according to the following formula: ; in, This represents the actual magnetic force output at time t. This indicates that the i-th magnetic particle unit is at a distance of and angle of action The magnetic gain coefficient under the following conditions This represents the distance between the i-th magnetic particle unit and the area of ​​action on the sole of the foot. This represents the angle between the magnetic pole direction of the i-th magnetic particle unit and the desired drag direction. This represents the current command of the i-th magnetic particle unit at time t. Let represent the magnetic pole direction control vector of the i-th magnetic particle unit at time t.

[0082] The magnetic gain coefficient is obtained from the factory calibration table, offline magnetic field simulation data, or platform operation calibration data. The target drag parameter vector is output from S4 by the control system. In this process, the effective drag coefficient is determined based on the current material label, gait status, and plantar normal force. And calculate the desired magnetic feedback: ; in, This represents the expected magnetic feedback at time t. The effective drag coefficient is represented by the value at time t. This represents the normal plantar force of the user at time t. This represents the user velocity estimation vector output by S3. This indicates the prevention of extremely small positive numbers with a denominator of 0.

[0083] The displacement cancels out the target velocity according to the following formula: ; in, This indicates that the platform displacement at time t cancels out the target velocity. This represents the displacement compensation coefficient. This represents the user velocity estimation vector output by S3.

[0084] The displacement compensation coefficient satisfies: ; in, This represents the displacement compensation coefficient. When... At that time, the platform completely offsets the user's displacement; when At that time, the platform performs partial displacement compensation.

[0085] The joint optimization objective function is: ; Where J represents the joint optimization objective function, Indicates the terrain simulation error weights. Indicates the resistance feedback deviation weight. Indicates the displacement compensation error weight. Indicates the smoothing weights of highly variable weights. Indicates the smoothing weight of current changes. Represents the target terrain elevation field. Indicates the actual height field of the platform. This indicates a desired magnetic feedback. Indicates the actual magnetic force output. This indicates that the platform displacement cancels out the target velocity. This indicates the displacement compensation speed that the platform can currently perform. This indicates the change in altitude between adjacent control cycles. This indicates the change in current between adjacent control cycles.

[0086] The constraints include the travel constraints of the liftable support unit: ; in, Indicates the minimum allowable height of the liftable support unit. Indicates the maximum permissible height of the liftable support unit. This indicates the height command for the j-th liftable support unit. This indicates the total number of liftable support units.

[0087] The current safety upper limit constraint is: ; in, This represents the current command for the i-th magnetic particle unit. Indicates the safe upper limit of current. This indicates the total number of magnetic particle units.

[0088] User security constraints are as follows: ; in, This represents the actual magnetic force output at time t. This indicates the user's security threshold.

[0089] The height change rate constraint is: ; in, This represents the height change of the j-th liftable support unit within an adjacent control cycle. This indicates the maximum permissible height change in a single period.

[0090] The current change rate constraint is: ; in, This represents the change in current of the i-th magnetic particle unit within an adjacent control cycle. This indicates the maximum permissible change in current per single cycle.

[0091] The control system defines the optimization variables as follows: ; Where z represents the joint optimization variable, and h represents the height command vector of the liftable support unit. This represents the current-pole direction control vector of the magnetic particle unit.

[0092] After linearizing or quadratically approximating the joint optimization objective function, we obtain the standard form of the quadratic programming: ; Where z represents the joint optimization variable, The matrix representing the quadratic terms of the optimization problem. Let represent the term vector of a linear problem.

[0093] The constraint form is: ; in, The constraint matrix represents the optimization problem. The constraint vector represents the optimization problem.

[0094] The control system adjusts based on the semantics of the XR scene and the user's motion state. , and When the terrain consists of steps, slopes, or rocks, improve... Prioritize ensuring the accuracy of terrain height simulation; when the material is ice, sand, or mud, improve... Prioritize ensuring realistic resistance feedback; when the user's walking speed is high or the remaining effective space on the platform is small, increase... Prioritize ensuring the effectiveness of displacement compensation. Upper and lower limits are set for each weight to prevent a single control target from becoming too large and causing other control targets to fail.

[0095] When the quadratic programming solver obtains a feasible solution within the preset solution time, the control system outputs the height command for the liftable support unit and the current-pole direction control vector for the magnetic particle unit. When the quadratic programming solution exceeds the control cycle, has no feasible solution, or has constraint conflicts, the control system enters a degraded control mode: maintaining the safety command of the previous cycle, reducing the magnetic field output according to the preset ramp decay function, and retracting the liftable support unit to the safe mid-height at a limited speed; if a feasible solution cannot be obtained for a consecutive preset number of cycles, the S6 safety protection process is triggered. The platform execution command output by S5 is used for S6 high-frequency closed-loop scheduling and execution feedback monitoring.

[0096] S6. The control system establishes a hierarchical closed-loop scheduling architecture based on a real-time operating system, allocating data acquisition, pressure field modeling, motion intent estimation, virtual terrain analysis, joint optimization solution, magnetic field update, safety monitoring, and online learning to tasks of different priorities. Data is transferred between tasks at each layer through a lock-free circular buffer. Each data packet contains a unified timestamp, data version number, sensor reliability, and verification identifier. Reading tasks only read the completely written data packets to avoid inconsistencies in control state caused by concurrent read / write operations.

[0097] The 1000Hz real-time layer performs magnetic field command updates, safety monitoring, and emergency stop protection. This layer reads the magnetic particle unit current-pole direction control vector output by S5 and updates the magnetic field output based on current change rate constraints and user safety force constraints. Safety monitoring objects include actuator current, magnetic force feedback, user balance status, height feedback of the liftable support unit, control command validity, and communication status.

[0098] When the current of any actuator continuously exceeds And lasting longer than the preset time When the current exceeds the limit, the control system determines that the magnetic force is abnormal; when the magnetic force feedback modulus exceeds the user's safety force threshold, the control system determines that the magnetic force is abnormal; when the control command is missing, the timestamp expires, or the data version number is discontinuous, the control system determines that the control link is abnormal.

[0099] The 500Hz control layer performs pressure field sampling, gait state updates, extended Kalman filter state updates, and magnetic field predictive control. This layer receives the unified sensing data stream output from S1 and the gait feature vector output from S2, and sends the motion intent parameters output from S3 to the safety monitoring logics S5 and S6. When the user's torso tilt angle exceeds a preset tilt angle threshold, or the pressure center of gravity continuously deviates from the current support polygon, or there is a sudden drop in plantar pressure accompanied by a rapid decrease in motion capture height, the control system determines that there is a risk of fall. The support polygon is determined by the outer contours of the current left and right foot contact areas.

[0100] The 100Hz strategy layer performs virtual terrain parsing, target shape generation, joint optimization solution, height command output for the liftable support unit, and data synchronization with the XR scene engine. This layer transmits the platform's actual height field, magnetic feedback, user estimated position, and abnormal states back to the XR scene engine, enabling the XR scene engine to adjust the virtual image, virtual motion speed, and interactive events according to the platform's actual state.

[0101] The 10Hz learning layer performs personalized parameter evaluation and candidate parameter updates. This layer reads historical execution feedback data and user feedback data, and only outputs control parameter adjustment suggestions, without directly sending execution commands to the liftable support unit or magnetic particle unit. The candidate parameters output by the learning layer must be verified by S5 joint optimization constraints and S6 safety constraints before they can be applied.

[0102] When the safety monitoring logic triggers the emergency stop protection, the control system stops receiving control commands that exceed safety constraints, reduces the magnetic particle array current to 0 or a safe holding value according to a preset ramp decay function, locks the liftable support unit in its current position or retracts it to a safe mid-height at a limited speed, sends a pause signal to the XR scene engine, and records the trigger reason, sensor status, control commands, and platform feedback logs. The control system will not re-enter normal control mode until the user confirms safety or the system self-test passes.

[0103] Output execution feedback data: ; in, This represents the execution feedback data at time t. Indicates the actual height field of the platform. Represents the magnetic feedback vector. Indicates terrain simulation error. Indicates resistance feedback error. This indicates the displacement compensation error. This indicates a safety status indicator. Execution feedback data is used for XR scene engine synchronization and S7 online learning and personalized adaptive optimization.

[0104] S7. Without altering the real-time control link from S1 to S6, the control system incorporates an online learning and personalized adaptive optimization module. This module employs a near-end strategy optimization algorithm or an equivalent constrained strategy optimization algorithm to perform long-term optimization of terrain simulation error weights, resistance feedback deviation weights, displacement compensation error weights, displacement compensation coefficients, and material-resistance mapping table parameters. The online learning and personalized adaptive optimization module only outputs parameter adjustment values ​​and does not directly output height commands, current commands, or magnetic pole direction control vectors for the liftable support unit.

[0105] The online learning state vector is defined as: ; in, This represents the online learning state vector at time t. This represents the gait feature vector output by S2. This represents the motion intent parameters output by S3. This represents the target shape data output by S4. This represents the execution feedback data output by S6. This indicates user feedback ratings.

[0106] Adjustment amount of output parameters for online learning module: ; in, This represents the amount of online learning parameter adjustment at time t. This represents the adjustment amount of the terrain simulation error weights. This indicates the amount of adjustment for the resistance feedback deviation weight. This indicates the adjustment amount of the displacement compensation error weight. This indicates the adjustment amount of the displacement compensation coefficient. This indicates the adjustment amount for the material-resistance mapping table parameters.

[0107] The reward function is defined as: ; in, This represents the learning reward value at time t. Indicates terrain simulation error. Indicates resistance feedback error. This indicates the displacement compensation error. Indicates the gait naturalness score, Indicates user comfort rating. , , , and These represent the reward weights for the corresponding evaluation items.

[0108] Gait naturalness score is calculated based on the symmetry of pressure distribution between the left and right feet, gait frequency stability, smoothness of the pressure center trajectory, and stability of support duration. User comfort score is obtained from user interaction feedback, sensory evaluation, or system-preset evaluation rules. The online learning module operates at a 10Hz learning layer, updating candidate parameters after accumulating a preset number of valid samples. The updated candidate parameters do not directly replace the current control parameters but are first added to the candidate parameter pool.

[0109] The control system performs safety boundary checks on candidate parameters. Candidate parameters can only be applied to the material-resistance mapping table in S4 or the joint optimization weights in S5 via atomic operations after satisfying constraints on the travel of the liftable support unit, the upper limit of current safety, the user safety force, the displacement compensation capability, and the upper and lower limits of the control weights. If, after application, the candidate parameters cause terrain simulation errors, resistance feedback errors, displacement compensation errors, or a deterioration in user comfort scores within a consecutive preset evaluation period, the control system automatically rolls back to the previous stable parameter set.

[0110] The control system supports parameter persistence across sessions. At the end of each use, the control system saves the user's gait statistics, stability control parameters, material preference parameters, and historical error indices. At the start of the next use, the control system reads the corresponding user's historical parameters as initial personalized parameters and continues to verify them through the real-time security links from S1 to S6.

[0111] Output a set of personalized parameters: ; in, Represents a set of user-personalized parameters. Indicates the terrain simulation error weights. Indicates the resistance feedback deviation weight. Indicates the displacement compensation error weight. This represents the displacement compensation coefficient. This represents the material-resistance mapping table. The set of personalized parameters is fed back to S4 and S5 for subsequent target shape generation and joint optimization control, thus forming a complete closed loop of "multi-source perception - pressure field modeling - motion intent estimation - virtual terrain analysis - joint optimization control - safe closed-loop execution - personalized parameter update".

[0112] Example 2: This example provides a control system for a magnetic particle array XR motion platform based on a flexible surface, including: The multi-source heterogeneous sensing data acquisition and spatiotemporal synchronization module is used to collect magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture posture and shoe sole magnetic feedback data. It performs time synchronization, coordinate transformation, anomaly verification and credibility weighting on the various data streams to generate a unified sensing data stream. The pressure field modeling and gait analysis module is used to establish a continuous pressure field based on pressure data in the unified sensing data stream, calculate the plantar pressure center of gravity, bipedal pressure ratio, plantar contact area and peak pressure, and analyze the gait state and gait feature vector. The motion intent estimation module is used to estimate the user's horizontal position, user velocity estimation vector, user gait scalar, user motion direction estimation value, and user displacement trend vector based on the unified perception data stream and gait feature vector, and generate motion intent parameters by using extended Kalman filtering. The virtual terrain parsing and target shape generation module is used to read the local terrain height map and material label map in the XR scene according to the motion intention parameters, complete the mapping between virtual coordinates and platform center coordinates, and generate target terrain height field, smooth target height set and target resistance parameter vector; The joint optimization control module is used to construct a joint optimization problem of terrain simulation, resistance feedback and displacement compensation based on motion intention parameters and target shape data. Under the constraints of stroke, current, safety force and rate of change, it solves the height command of the liftable support unit and the current-magnetic pole direction control vector of the magnetic particle unit. The high-frequency closed-loop scheduling and safety protection module is used to perform hierarchical closed-loop scheduling of platform execution instructions based on the real-time operating system, perform magnetic field updates, support altitude control, safety monitoring, anomaly degradation and emergency stop protection, and output execution feedback data; The online learning and personalized adaptive optimization module is used to update the personalized parameter set based on gait feature vectors, motion intention parameters, target shape data, execution feedback data and user feedback scores, and feed the personalized parameter set back to the virtual terrain parsing and target shape generation module and the joint optimization control module.

[0113] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.

[0114] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0115] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0116] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A control method for an XR motion platform based on a flexible surface magnetic microparticle array, characterized in that, Includes the following steps: S1. Collect magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture posture and shoe sole magnetic feedback data, and perform time synchronization, coordinate transformation, anomaly verification and credibility weighting on each data source to generate a unified perception data stream; S2. Establish a continuous pressure field based on the pressure data in the unified sensing data stream, calculate the plantar pressure center of gravity, bipedal pressure ratio, plantar contact area and peak pressure, and analyze to obtain gait state and gait feature vector; S3. Based on the unified perception data stream and gait feature vector, the extended Kalman filter is used to estimate the user's horizontal position, user velocity estimation vector, user gait scalar, user movement direction estimation value, and user displacement trend vector to generate motion intention parameters. S4. Read the local terrain height map and material label map in the XR scene according to the motion intention parameters, complete the mapping between virtual coordinates and platform center coordinates, and generate the target terrain height field, smooth target height set and target resistance parameter vector. S5. Based on the motion intention parameters and target shape data, construct a joint optimization problem of terrain simulation, resistance feedback and displacement compensation, and solve the height command of the liftable support unit and the current and magnetic pole direction control vector of the magnetic particle unit under the constraints of stroke, current, safety force and rate of change. S6. Based on the real-time operating system, the platform executes instructions in a hierarchical closed-loop scheduling manner, performs magnetic field updates, supports altitude control, safety monitoring, anomaly degradation and emergency stop protection, and outputs execution feedback data.

2. The control method for the XR motion platform of a magnetic particle array based on a flexible surface according to claim 1, characterized in that, Also includes: S7. Update the personalized parameter set based on gait feature vectors, motion intention parameters, target morphology data, execution feedback data, and user feedback scores, and feed the personalized parameter set back to the target morphology generation and joint optimization control process.

3. The control method for the XR motion platform of a magnetic particle array based on a flexible surface according to claim 1, characterized in that, S1 specifically includes: A platform center coordinate system was established with the geometric center of the platform's flexible surface, and magnetic particle unit pressure, bipedal inertial measurement data, optical motion capture posture, and shoe sole magnetic feedback data were collected respectively. Based on hardware timestamps, the data from each channel is aligned to a unified time axis, and the data from each channel is transformed to the platform's central coordinate system using rotation matrices and translation vectors. The system verifies timestamps, pressure jumps, inertial measurement anomalies, motion capture occlusion, and magnetic feedback anomalies, and calculates sensor reliability weights based on the anomaly errors to generate a unified sensing data stream.

4. The control method for the XR motion platform of a magnetic particle array based on a flexible surface according to claim 1, characterized in that, S2 specifically includes: A continuous pressure field is established by bilinear interpolation based on the pressure output value and geometric position of the magnetic particle unit; The centroid of plantar pressure, the sum of pressure on the left foot, the sum of pressure on the right foot, the ratio of pressure on both feet, the plantar contact area, and the peak pressure are calculated based on the continuous pressure field. Based on the bipedal pressure ratio, plantar contact area, peak pressure, and pressure change rate, the system identifies the right foot support state, left foot support state, bipedal support state, foot lift state, and foot landing state, and generates gait feature vectors.

5. The control method for the XR motion platform of a magnetic particle array based on a flexible surface according to claim 1, characterized in that, S3 specifically includes: An extended Kalman filter state vector is constructed using the user's horizontal position, horizontal velocity vector, forward direction angle, and inertial measurement unit acceleration bias. The pressure center of gravity displacement observation, optical motion capture hip joint position observation, and inertial measurement short-time integral displacement observation are combined into a joint observation vector; Adjust the observation noise covariance and process noise covariance based on the sensor confidence weight and gait state, and output the user velocity estimation vector, user gait scalar, user motion direction estimation value and user displacement trend vector to form motion intention parameters.

6. The control method for the XR motion platform of a magnetic microparticle array based on a flexible surface according to claim 1, characterized in that, S4 specifically includes: Read local terrain height maps, material label maps, and scene semantic labels through the XR scene interface; The mapping relationship between XR virtual coordinates and platform center coordinates is established based on the virtual-to-real ratio coefficient and the rotation alignment matrix; Based on the user's horizontal position, estimated user movement direction, and user walking speed scalar, the local virtual terrain corresponding to the current foot contact area and the predicted foot landing area is extracted, and a smooth target height is generated for each liftable support unit; Based on the material label of the sole contact area, query the material and resistance mapping table to generate the target resistance parameter vector.

7. The control method for the XR motion platform of a magnetic microparticle array based on a flexible surface according to claim 1, characterized in that, S5 specifically includes: Construct the actual height field of the platform based on the target terrain height field and the height instructions of the liftable support unit; The actual magnetic output is calculated based on the current command of the magnetic particle unit, the magnetic pole direction control vector, the magnetic gain coefficient, and the position of the action area on the sole of the foot. The desired magnetic feedback is calculated based on the target resistance parameter vector, the user's normal foot force, and the user's velocity estimation vector. A joint optimization objective is constructed using terrain simulation error, resistance feedback deviation, displacement compensation error, height change smoothing term, and current change smoothing term. The platform execution command is then solved under constraints of travel, current, safety force, height change rate, and current change rate.

8. The control method for the XR motion platform of a magnetic particle array based on a flexible surface according to claim 1, characterized in that, S6 specifically includes: Based on the real-time operating system, a 1000Hz real-time layer, a 500Hz control layer, a 100Hz strategy layer, and a 10Hz learning layer are set up. The 1000Hz real-time layer performs magnetic field command updates, safety monitoring, and emergency stop protection; the 500Hz control layer performs pressure field sampling, gait state updates, and extended Kalman filter state updates; the 100Hz strategy layer performs virtual terrain parsing, joint optimization solutions, and XR scene engine data synchronization; and the 10Hz learning layer performs personalized parameter evaluation. When there is an overcurrent, abnormal magnetic force, risk of falling, or abnormal control link, the system will perform magnetic field attenuation, support height safe retreat, and XR scene pause.

9. The control method for the XR motion platform of a magnetic particle array based on a flexible surface according to claim 2, characterized in that, S7 specifically includes: The gait feature vector, motion intention parameters, target shape data, execution feedback data, and user feedback scores are used to construct the online learning state vector; A reward function is constructed using terrain simulation error, resistance feedback error, displacement compensation error, gait naturalness score, and user comfort score. Output terrain simulation error weights, resistance feedback deviation weights, displacement compensation error weights, displacement compensation coefficients, and parameter adjustments for material and resistance mapping tables; After performing safety boundary checks and degradation rollbacks on the candidate parameters, a personalized parameter set is formed and fed back to the target morphology generation and joint optimization control process.

10. A control system for a magnetic particle array XR motion platform based on a flexible surface, employing the control method for a magnetic particle array XR motion platform based on a flexible surface as described in any one of claims 1 to 9, characterized in that, include: The system includes a multi-source heterogeneous sensing data acquisition and spatiotemporal synchronization module, a pressure field modeling and gait analysis module, a motion intent estimation module, a virtual terrain analysis and target shape generation module, a joint optimization control module, a high-frequency closed-loop scheduling and safety protection module, and an online learning and personalized adaptive optimization module.

Images