A plantar and ankle force detection method based on a six-dimensional force sensor

Abstract

The application discloses a kind of based on six-dimensional force sensor's foot bottom and ankle stress detection method, steps are as follows: S1, three sensors are used to measure respective force and moment of force respectively;S2, data fusion is carried out, and fusion model is constructed;S3, extract multiple key parameters, including foot bottom surface support surface attitude angle, ground friction direction and size, forefoot and hindfoot stress distribution and change rate, support point position and foot center offset, are applied to identification support surface stability, detection slip trend, identification gait phase, real-time adjustment foot bottom attitude or ankle output angle in sequence respectively.The application is extracted from data fusion by all-around six-dimensional force detection, respectively from foot, forefoot, hindfoot and ankle, and then based on fusion model estimation current state is adjusted specifically, effectively constitutes three-point arrangement, full-dimensional force detection, data fusion, comprehensive perception scheme of support control feedback.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot perception and control, and particularly relates to a plantar and ankle force detection method based on a six-dimensional force sensor. BACKGROUND

[0002] In application scenarios such as humanoid robots, lower limb exoskeleton systems, and rehabilitation gait assistance devices, foot-ground contact state perception is a core technology for realizing stable gait control, real-time posture adjustment, fall prediction, and environmental interaction. Existing plantar force perception systems are mainly divided into three categories. The first category is a perception system based on pressure distribution, such as embedding multiple single-axis pressure sensors into the plantar structure to collect normal pressure data in different areas of the plantar. This type of solution has the advantages of simple structure and low cost, but has the problems of missing dimension information, large force transmission path error, and inability to accurately reconstruct the contact posture. This type of system only detects normal pressure and cannot obtain tangential force, friction torque, or torque. The force from the contact surface to the sensor is coupled through flexible structures, which easily introduces errors, and it is also difficult to determine the roll angle, pitch angle, and ground friction direction of the support area using multiple single-axis pressure combination systems.

[0003] The second category is a scheme based on flexible film and capacitive array, which records the plantar pressure distribution map through a flexible film and a sensing array, and is suitable for rehabilitation training and gait analysis. The advantages of this scheme are that it is wearable and can be arranged in a large area, but it also has obvious shortcomings. It lacks three-dimensional force perception capability and cannot detect shear force or torsional torque. There are obvious noise interference and temperature drift phenomena, and the data is unstable. Moreover, this scheme relies on thin film materials, which have limited material life and poor fatigue resistance, and is not suitable for robots to be used repeatedly for a long time. The third category is a scheme that partially uses six-dimensional force sensors but is not reasonably arranged. Existing research has attempted to use six-dimensional force sensors in robot feet to measure single-point force and torque, which is often arranged at the ankle joint. This scheme cannot cover the entire plantar support surface, resulting in inaccurate estimation of support point deviation, inability to determine independent force changes in the forefoot and hindfoot, and lack of a systematic fusion modeling method for control feedback force-posture inference. This type of solution has high cost and low data utilization.

[0004] As can be seen, in the prior art, the current robot foot force perception system at least has the following problems:

[0005] 1. Insufficient perception dimension: most existing plantar sensing schemes can only detect vertical pressure and cannot obtain complete three-axis force and three-axis torque, resulting in incomplete understanding of ground reaction force (GRF) and difficulty in estimating friction direction and contact posture;

[0006] 2. The perception area is not fully covered: sensors are usually arranged centrally on the foot or below the ankle, ignoring the independent force differences of the forefoot and hindfoot, and cannot determine the landing sequence and local support situation;

[0007] 3. Lack of data fusion modeling processing: existing systems output raw torque data multiple times, lack a unified fusion framework, and cannot accurately estimate the foot support surface, support point position (COP), ground friction direction, and support surface pitch angle / roll attitude;

[0008] 4. Cannot be applied to control closed-loop feedback: controllable mapping between sensor data and foot state estimation is not achieved, making it difficult to directly use in robot stable walking, center of gravity adjustment, and other controls. SUMMARY

[0009] To solve the above four problems, the purpose of the present application is to propose a foot and ankle force detection method based on a six-dimensional force sensor, which extracts data from the foot, forefoot, hindfoot, and ankle joint for fusion through comprehensive six-dimensional force detection, and then estimates the current state based on the fusion model for targeted adjustment, effectively forming a comprehensive perception scheme of three-point arrangement, full-dimensional force detection, data fusion, and support control feedback.

[0010] The technical scheme is as follows:

[0011] A foot and ankle force detection method based on a six-dimensional force sensor, using a robot foot for detection, comprising the following steps:

[0012] S1, using a first sensor to detect the force and torque between the ankle joint and the foot platform, simultaneously measuring the vertical load change of the whole foot and the overall torque output under the current attitude of the ankle joint; using a second sensor to obtain the landing instantaneous impact, friction force change, and support end force corresponding to the forefoot; using a third sensor to obtain the landing instantaneous impact, friction force change, and support end force corresponding to the hindfoot;

[0013] S2, using the data of the three sensors in step S1 for data fusion, constructing a fusion model based on a foot biomechanical model, and the structure of the fusion model adopts a biological three-point support distribution of the ankle, forefoot, and hindfoot, for simulating the force conduction path of the arch structure;

[0014] S3, extracting multiple key parameters from the fusion model of step S2, including the foot surface support surface attitude angle, the direction and size of the ground friction force, the force distribution and change rate of the forefoot and hindfoot, and the offset amount of the support point position and the foot center, which are used in turn to identify support surface stability, detect slip trend, identify gait phase, and adjust foot attitude or ankle output angle in real time.

[0015] Preferably, the robot foot includes a sole and an ankle joint. The sole is an asymmetrical elongated ellipse and includes a forefoot and a heel. Both the forefoot and the heel are connected to pads and cushioning pads.

[0016] Preferably, in step S1, the three sensors are all six-dimensional force sensors of the same model; each sensor is equipped with a standard interface to support at least the controller local area network communication protocol and the RS485 communication protocol; each standard interface is connected to an embedded processor for data acquisition and real-time processing.

[0017] Preferably, in step S1, the first sensor is fixed at the center of the ankle joint, the second sensor is fixed at the center of the forefoot, and the third sensor is fixed at the center of the heel. Each sensor is fixed by multiple bolts, and the second and third sensors are kept on the same horizontal plane.

[0018] Preferably, the data fusion process in step S2 is as follows:

[0019] 1) Based on the data from each sensor and their respective spatial distribution on the sole of the foot, construct each force-torque coupling matrix G. i This is used to obtain the resultant force and resultant torque F of each sensor at the overall reference point O1 on the foot. REF ;

[0020] 2) The weighted least squares method based on the biomechanical weight distribution of the foot is used to fuse the data from each sensor;

[0021] 3) Introduce extended Kalman filtering to filter the fusion result, predict the next state, and correct errors;

[0022] 4) Extract multiple feature indicators in real time from the results of the error correction, including the location of the support point, the location of the zero moment point, and the direction of the resultant force.

[0023] Preferably, each force-moment coupling matrix G is constructed. i First, a foot coordinate system is constructed with the ankle joint as the origin and the vertical upward axis as the z-axis. The y-axis of this system points perpendicularly from the origin to the line connecting the centers of the second and third sensors, and the x-axis is perpendicular to the y-axis. Then, the data from each sensor is recorded as follows: , These correspond to the ankle joint, forefoot, and heel, respectively. T represents transpose, f represents force, and m represents torque. The subscripts x, y, and z correspond to the three axes of the coordinate system. ,in, Let R represent the set of real numbers, where R represents the set of real numbers. 6 R represents a column vector with 6 combinations of real numbers and is read as a 6-dimensional real vector space. 6×6represents a 6x6 real matrix space, represents a 3D identity matrix, which is used to keep the original force direction, represents a 3D zero matrix, which indicates that there is no direct coupling between force and moment, r i for representing the position-induced moment influence, is the inverse matrix of r i , ; the resultant force and the resultant moment of each sensor at the global reference point O1 , where F total represents the resultant force of the force vectors of each sensor after being uniformly converted to the global reference point O1, M total represents the total moment of the center of each sensor to the global reference point O1.

[0024] Preferably, the formula for data fusion is: , where is a weight matrix, and T represents transposition, is the weight corresponding to each of the three sensors, and diag represents the form of a diagonal matrix, is the combination of the total coupling matrix of the three sensors, is the data combination of the forces of the three sensors.

[0025] Preferably, when multiple feature indicators are extracted in real time, the generalized force on the foot after data fusion is represented as , and the generalized force X acts on the global reference point O1; wherein , , respectively correspond to the components of the force on the x-axis, y-axis, and z-axis after data fusion, , , respectively correspond to the moment components opposite to , , ; the zero moment point is set on the ground and has a two-dimensional coordinate of ; wherein , , h is the vertical distance from the global reference point O1 to the foot, and ZMP represents the zero moment point.

[0026] Preferably, the gait phase in step S3 includes the initial contact phase, the mid-stance phase, and the push-off phase; according to the current gait phase and the slip trend, the pitch angle and the roll angle of the ankle joint are adjusted, and the gait is corrected.

[0027] Preferably, before adjusting the foot posture in step S3, the pitch angle estimation and the roll angle estimation of the posture are performed, including: performing angle estimation based on the deviation of the resultant force direction relative to the z-axis; setting the pitch angle is the inclination angle of the y-axis rotation, and the inclination angle of the pitch angle is determined based on the component force in the x-axis direction; the roll angle is set is the inclination angle of the x-axis rotation, and the inclination angle of the roll angle is determined based on the component force in the y-axis direction; wherein, , arctan represents an inverse tangent function, represents foot inclination, represents foot inclination, represents foot inclination, represents foot inclination; based on the results of the pitch angle estimation and the roll angle estimation, the foot posture is adjusted.

[0028] Compared with the prior art, the present application has the beneficial effects that:

[0029] I) Realize all-around six-dimensional force detection, and improve the integrity of the perception dimension: the six-dimensional force sensor has high perception accuracy, three six-dimensional force sensors are arranged to cover the forefoot, the hindfoot and the ankle area, the three-axis force and the three-axis moment of each measuring point can be detected at the same time, the direction, size and distribution of the ground reaction force are perceived in real time, the defects that the traditional piezoelectric and strain gauge sensors can only obtain vertical force or limited direction are made up, and therefore the spatial integrity and accuracy of the foot contact perception are significantly improved;

[0030] II) Cover the key area of the foot, support local force analysis and gait phase recognition: the three-point arrangement mode of the present application takes into account the forefoot propulsion area, the hindfoot buffer area and the ankle joint main support area, the moment changes of each part can be independently detected, the robot landing sequence (the forefoot lands first or the hindfoot lands first), the support stability (whether the center of gravity is shifted) and the current phase (touching the ground, supporting, leaving the ground) can be judged, the accurate division of the support phase and the swing phase in the gait cycle is realized, and the basis for stable control and gait planning is provided;

[0031] III) Estimate key support parameters based on the fusion model of mechanics, and improve the control ability of the robot: the model obtained by fusing the force and the moment can calculate the support point position (COP) and the support surface posture angle (pitch angle , roll angle ), the offset control of the support point position, the compensation control of the foot inclination surface, the dynamic support adjustment strategy, the closed-loop control and the like can be realized, and the real-time perception and adjustment of the stability of the lower limbs of the robot, the center of gravity adjustment and the posture maintenance are realized;

[0032] IV) Compact structure, easy to integrate, and convenient for standard robot system popularization and application: all the sensors are arranged compactly, adapt to the standard foot structure, the support structure adopts a modular design, is convenient for installation, disassembly and replacement, has good compatibility and engineering practicability, and is suitable for engineering practice. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 This is a flowchart of a method for detecting force on the sole and ankle based on a six-dimensional force sensor;

[0034] Figure 2 A schematic diagram of the installation of a three-point six-dimensional force sensor on the sole of the foot;

[0035] Figure 3 This is a schematic diagram illustrating the synthesis of force and torque in the foot.

[0036] Figure 4 A block diagram of a foot detection system;

[0037] Figure 5 This is a flowchart of a posture recognition and control process.

[0038] Figure labels: 1 First sensor, 2 Second sensor, 3 Third sensor, 4 Ankle joint, 5 Heel of foot, 6 Forefoot. Detailed Implementation

[0039] The following will be combined with the present invention. Figures 1 to 5 The technical solutions in the embodiments of the present invention will be described in detail below.

[0040] like Figure 1 The diagram shows a flowchart of a foot and ankle force detection method based on a six-dimensional force sensor. Through comprehensive six-dimensional force detection, data is extracted from the forefoot (6), heel (5), and ankle (4) and fused. Based on the fusion model, the current state is estimated and targeted adjustments are made. This effectively constitutes a comprehensive perception scheme with three-point arrangement, full-dimensional force detection, data fusion, and support for control feedback.

[0041] The specific method is as follows:

[0042] S1. The first sensor 1 is used to detect the force and torque between the ankle joint 4 and the foot platform, and at the same time, the vertical load change of the whole foot and the overall torque output by the ankle joint 4 in the current posture are measured; the second sensor 2 is used to obtain the instantaneous impact, friction change and support end force of the forefoot 6; the third sensor 3 is used to obtain the instantaneous impact, friction change and support end force of the heel.

[0043] like Figure 2The diagram shows a schematic of a three-point, six-dimensional force sensor installation on the sole of the foot. In the diagram, 1 represents the first sensor, 2 represents the second sensor, 3 represents the third sensor, 4 represents the ankle joint, 5 represents the heel, and 6 represents the forefoot. The robot's foot mainly includes the sole, ankle joint 4, and multiple connecting components between them. The sole is an asymmetrical elongated ellipse, comprising the forefoot 6 and heel 5. Both the forefoot 6 and heel 5 are connected to pads and cushioning pads, which then support the robot on the ground. This three-point arrangement takes into account the forefoot propulsion zone, the heel buffer zone, and the main support zone of the ankle joint 4. It allows for independent detection of torque changes in each part, enabling subsequent determination of the robot's landing sequence (whether the forefoot 6 or heel 5 lands first), support stability (whether the center of gravity has shifted), and the current stage (touch, support, lift-off). This allows for precise division of the support and swing phases in the gait cycle, providing a basis for stability control and gait planning.

[0044] In this embodiment, all three sensors are six-dimensional force sensors of the same model. Each sensor is equipped with a standard interface for data transmission and supports at least the Controller Area Network (CAN) communication protocol and the RS485 communication protocol. Each standard interface is connected to an embedded processor for data acquisition and real-time processing. The first sensor 1 is fixed at the center of the ankle joint 4, the second sensor 2 is fixed at the center of the forefoot 6, and the third sensor 3 is fixed at the center of the heel 5. Each sensor is fixed by multiple bolts, and the second sensor 2 and the third sensor 3 are kept on the same horizontal plane.

[0045] Specifically, three six-dimensional force sensors are respectively installed in the middle of the forefoot, the middle of the heel, and below the center of the ankle joint 4. The total length of the foot is 245mm. The first sensor 1 is located 30mm behind the center of the foot and 70mm above the foot, below the center of the ankle joint 4. The second sensor 2 is located 80mm in front of the center of the foot in the middle of the forefoot. The third sensor 3 is located 40mm from the end of the heel near the center of the foot. Each sensor has its force transmission surface below it and is fixed with bolts. It is connected to the ground by a gasket and a buffer pad. The lines connecting the three sensors form a stable triangular layout structure to achieve reliable force transmission.

[0046] S2. Using the data from the three sensors in step S1, data fusion is performed to construct a three-point torque fusion model based on foot biomechanics. The structure of the fusion model adopts a three-point support distribution of the ankle joint, forefoot 6 and heel 5 to simulate the force transmission path of the foot arch structure.

[0047] like Figure 3 The diagram shown illustrates the synthesis of foot force and torque. In the diagram, A, Q, and L correspond to the centers of the first sensor 1, the second sensor 2, and the third sensor 3, respectively. O1 represents the overall reference point for each sensor on the foot.Figure 4 The diagram shown is a block diagram of a foot detection system; combined with Figure 3 and Figure 4 As shown, the following process is required when synthesizing fused data:

[0048] 1) Based on the data from each sensor and their respective spatial distribution on the sole of the foot, construct each force-torque coupling matrix G. i This is used to obtain the resultant force and resultant torque F of each sensor at the overall reference point O1. REF .

[0049] Specifically, in constructing each force-moment coupling matrix G i First, a foot coordinate system is constructed with the ankle joint (4) as the origin and the vertical upward axis as the z-axis. The y-axis of this system points perpendicularly from the origin to the line connecting the centers of the second sensor (2) and the third sensor (3), and the x-axis is perpendicular to the y-axis. Then, the data from each sensor is recorded as follows: , 4 corresponds to the ankle joint, 6 to the forefoot, and 5 to the heel, respectively. T represents transpose, f represents force, and m represents torque. The subscripts x, y, and z correspond to the three axes of the coordinate system.

[0050] Then, a coupling matrix G is constructed based on the data from each sensor. i , ,in, Let R represent the set of real numbers, where R represents the set of real numbers. 6 F represents a column vector consisting of 6 combinations of real numbers and is read as a 6-dimensional real vector space. i It is a six-dimensional column vector, R 6×6 This represents a matrix with 6 rows and 6 columns, and is read as a 6×6 real matrix space. Represents a 3D identity matrix, i.e. The 3D identity matrix is ​​used to maintain the direction of the original force. Represents a 3D zero matrix, i.e. A 3D zero matrix indicates that there is no direct coupling between force and torque, r i Used to characterize the effect of position on torque. For r i The anti-matrix, Continue to obtain the resultant force and resultant torque of each sensor at the overall reference point O1. , of which F total M represents the resultant force after the force vectors of all sensors are uniformly converted to the overall reference point O1. total This represents the total torque of the centers of each sensor relative to the overall reference point O1.

[0051] 2) Using the weighted least squares method based on foot biomechanics weight distribution, the data of each sensor is processed with different weights by least squares method, so as to complete data fusion.

[0052] Specifically, the formula of data fusion is: , wherein, is a weight matrix, T represents transposition, is the weight corresponding to the first sensor 1, the second sensor 2 and the third sensor 3 respectively, diag represents the form of diagonal matrix, is the total coupling matrix combination of the three sensors, is the data combination of the three sensors.

[0053] 3) Introducing extended Kalman filter EKF, the real-time state estimation and noise suppression of the fusion result is carried out, the error is effectively corrected, and the data accuracy is improved.

[0054] It should be noted that EKF is the full name of Extended Kalman Filter, that is, extended Kalman filter, which is an algorithm for processing state estimation problems of nonlinear systems. By Taylor expansion of nonlinear function and ignoring high-order terms, the first-order linearization part is retained to approximate the original nonlinear system, so that the standard Kalman filter can be applied to nonlinear systems, which can effectively correct errors.

[0055] 4) From the result of completing error correction, a plurality of characteristic indexes can be extracted in real time, including support point position, zero moment point position and resultant force direction. When a plurality of characteristic indexes are extracted in real time, the generalized force on the foot after data fusion is recorded as , and the generalized force X acts on the overall reference point O1; wherein, , , respectively correspond to the force on the x axis, y axis and z axis after data fusion, , , respectively correspond to the force moment opposite to , , . Then, the zero moment point is set on the ground and the two-dimensional coordinates are , wherein, , , h is the vertical distance from the overall reference point O1 to the foot, and ZMP represents the zero moment point.

[0056] S3. Extract several key parameters from the fusion model in step S2, including the plantar support surface posture angle, the direction and magnitude of ground friction, the force distribution and rate of change of the forefoot 6 and heel 5, and the offset of the support point position from the foot center. These parameters are then applied sequentially to identify support surface stability, detect slippage trends, identify gait phases, and adjust the plantar posture or ankle joint 4 output angle in real time. Gait phases include the initial contact phase, the mid-support phase, and the push-off phase. The initial contact phase is the instant the heel 5 begins to contact the ground, marking the beginning of a gait cycle, with the overall center of gravity gradually shifting towards the foot position that just made contact. The mid-support phase occurs after the entire plantar surface has fully contacted the ground, at which point the weight is evenly distributed on the standing leg. The push-off phase occurs during the forward movement of the overall center of gravity, at which point the forefoot 6 prepares to leave the ground and enter the swing phase to generate thrust in preparation for the next gait cycle.

[0057] Specifically, when extracting multiple key parameters from the model, the direction of the resultant force is first calculated, which is the resultant force vector. After further normalization, it becomes the direction vector. This is the direction cosine vector of the spatial resultant force, which can be used to estimate the normal of the support surface, i.e., to judge the attitude tilt trend, dynamically estimate the direction of the center of gravity offset, calculate the friction control, etc., and then it can be used to identify the stability of the support surface, detect the slip trend, identify the gait stage, and adjust the foot posture or the output angle of the ankle joint in real time.

[0058] like Figure 5 The diagram shows a flowchart of a posture recognition and control process. First, based on the current gait stage and sliding trend, the current posture is determined, and the pitch and roll angles of the ankle joint 4 are estimated to confirm the plantar support posture angle. Then, combined with the known support point position and plantar support posture angle, it can be determined whether the landing is forefoot-first or heel-first, and whether the current gait is in the take-off phase, support phase, or initial contact phase. Finally, the pitch and roll angles of the ankle joint 4 are adjusted to correct the gait and adjust the plantar posture.

[0059] When estimating pitch and roll angles, first estimate the angle based on the offset of the resultant force direction relative to the z-axis; then set the pitch angle. The tilt angle is the rotation along the y-axis, and the pitch angle is determined based on the component of force along the x-axis; the roll angle is set. The tilt angle is the x-axis, and the magnitude of the roll angle is determined based on the component of the force along the y-axis; where, , arctam represents the arctangent function. Indicates that the feet are tilted forward. Indicates the backward extension of the feet; when represents foot pronation, and represents foot supination. Then, when the two inclination angles do not match the current gait phase, the pitch angle and roll angle of the posture can be adjusted so that the foot posture can match the gait phase.

[0060] In summary, the application adopts a high-precision six-dimensional force sensor and covers the forefoot, hindfoot and ankle areas in a three-point arrangement to realize all-around force and torque detection, significantly improving the integrity and accuracy of the perception dimension. This arrangement can not only obtain the direction, size and distribution of the ground reaction force in real time, but also make up for the deficiency of the traditional sensor that can only detect limited directional forces. Moreover, this design supports accurate division of each stage of gait and local force analysis, providing more stable support for foot movement analysis and gait planning basis. In addition, based on the fusion model of mechanics, key support parameters such as support point position and posture angle can be estimated, thereby enhancing the adjustment control ability of the foot, which has significant progressiveness.

[0061] The above examples only illustrate the technical idea of the application, and cannot limit the protection scope of the application. Any modification made according to the technical idea of the application on the basis of the technical scheme falls within the protection scope of the application.

Claims

1. A method for detecting plantar and ankle force based on a six-dimensional force sensor, using a robotic foot for detection, characterized by, Comprise the following steps: S1, using the first sensor detects the force and moment between the ankle joint and the foot platform, while measuring the whole foot vertical load change and the overall moment output under the current attitude of the ankle joint; using the second sensor to obtain the corresponding landing instantaneous impact, friction force change and support end force of the forefoot; Using the third sensor to obtain the corresponding landing instantaneous impact, friction force change and support end force of the hindfoot; S2, using the data of the three sensors in step S1 to perform data fusion, construct a fusion model based on the foot biomechanical model, the structure of the fusion model adopts the biological three-point support distribution of the ankle, forefoot and hindfoot, which is used to simulate the arch structure force transmission path; the process of data fusion is as follows: 1) building each force-moment coupling matrix G from the data of each sensor and the respective spatial distribution position on the plantar surface i , for obtaining the resultant force and moment F REF of each sensor at the global reference point O1 of the foot; Constructing each force-moment coupling matrix G i When, first with the ankle joint as the origin, vertically to the ground upward as the z axis to construct the foot coordinate system, wherein the y axis of the foot coordinate system is vertically directed to the center line of the second sensor and the third sensor from the origin, and the x axis is perpendicular to the y axis; Then record the data of each sensor as , Corresponding to the ankle joint, forefoot, and hindfoot, respectively, T represents transposition, f represents force, m represents moment, and the subscripts x, y, and z correspond to the three axes of the coordinate system; wherein , wherein Belongs to, R represents the set of real numbers, R 6 Indicates a column vector with 6 real number combinations and is read as a 6-dimensional real number vector space, R 6×6 Indicates a 6x6 matrix and is read as a 6x6 real matrix space, Indicates a 3-dimensional unit matrix, and the 3-dimensional unit matrix is used to maintain the original force direction, Indicates a 3-dimensional zero matrix, and the 3-dimensional zero matrix indicates that there is no direct coupling between force and moment, i Used to represent the position-induced moment effect, Is the inverse matrix of r i , ; The total force and total moment of each sensor at the overall reference point O1 are , wherein F total Indicates the total force after the force vector of each sensor is uniformly converted to the overall reference point O1, and M total Indicates the total moment of the center of each sensor to the overall reference point O1; 2) using weighted least squares method based on foot biomechanical weight distribution to fuse the data of each sensor; The formula of data fusion is: wherein, is a weight matrix, T represents transposition, is the weight corresponding to the three sensors respectively, diag represents the form of diagonal matrix, is the total coupling matrix combination of the three sensors, is the data combination of the three sensors forces; 3) introducing extended Kalman filter to filter the fusion results and predict the next state and correct the error; 4) from the results of completing the error correction, real-time extraction of multiple characteristic indexes, including support point position, zero moment point position and resultant force direction; When the plurality of feature indexes are extracted in real time, the generalized force on the foot sole after data fusion is denoted as , and the generalized force X acts on the overall reference point O1; wherein, , , respectively correspond to the force components on the x-axis, y-axis and z-axis after data fusion, , , respectively correspond to the moment components relative to , , ; the zero moment point is set on the ground and has a two-dimensional coordinate ; wherein, , , h is the vertical distance from the overall reference point O1 to the foot sole, and ZMP represents the zero moment point. S3, extract multiple key parameters from the fusion model of step S2, including foot bottom support surface attitude angle, ground friction force direction and size, forefoot and hindfoot stress distribution and change rate, support point position and foot center offset, which are used to identify support surface stability, detect slip trend, identify gait phase, real-time adjust foot bottom attitude or ankle output angle respectively.

2. The plantar and ankle force detection method based on the six-dimensional force sensor according to claim 1, wherein, The robot foot comprises a foot bottom and an ankle joint, the foot bottom is asymmetrically long elliptical, the foot bottom comprises a forefoot and a hindfoot, and the forefoot and the hindfoot are connected with a gasket and a buffer pad.

3. The method of claim 1, wherein the six-axis force sensor is used to measure the force applied to the foot and the ankle. In step S1, the three sensors are all unified six-dimensional force sensors; each sensor is provided with a standard interface for supporting at least controller area network communication protocol and RS485 communication protocol; each standard interface is connected to an embedded processor for data acquisition and real-time processing.

4. The plantar and ankle force detection method based on the six-dimensional force sensor according to claim 1, wherein, In step S1, the first sensor is fixed at the center of the ankle joint, the second sensor is fixed at the center of the forefoot, and the third sensor is fixed at the center of the hindfoot, each sensor is fixed through a plurality of bolts, and the second sensor and the third sensor are kept in the same horizontal plane.

5. The method of claim 1, wherein the six-axis force sensor is used to measure the force applied to the foot and the ankle. The gait phase in step S3 comprises an initial contact phase, a mid-stance phase and a push-off phase; according to the current gait phase and the slip trend, the pitch angle and the roll angle of the ankle joint are adjusted and the gait is corrected.

6. The method of claim 1, wherein the six-axis force sensor is used to measure the force applied to the foot and the ankle. Before adjusting the foot posture in step S3, the pitch angle estimation and the roll angle estimation of the posture are performed, including: performing angle estimation based on the offset of the resultant force direction relative to the z-axis; setting the pitch angle is the inclination angle of y-axis rotation, and the inclination size of the pitch angle is determined based on the component force in the x-axis direction; setting the roll angle is the inclination angle of x-axis rotation, and the inclination size of the roll angle is determined based on the component force in the y-axis direction; wherein, , , arctan denotes the inverse tangent function, when denotes a foot anteversion, denotes a foot retroversion; when denotes a foot varus, , denotes a foot valgus; the foot posture is adjusted based on the results of the pitch angle estimation and the roll angle estimation.

Images