A method and system for detecting foot end contact force and position of a wheel-legged robot

By setting up multiple pressure sensors on the tire and establishing corresponding models, the problem of wheel-leg robot sensing contact state in complex terrain is solved, accurate detection of foot contact force and position is achieved, and the travel efficiency and stability of the robot are improved sex.

CN118636196BActive Publication Date: 2024-12-06WUHAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411111473.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-12-06
Estimated Expiration
2044-08-14

AI Technical Summary

Technical Problem

Existing wheel-leg robots are difficult to accurately sense the foot-end contact state and contact force in complex unstructured terrain, and cannot meet the foot-end shape detection and force detection requirements of wheel-leg movement.

Method used

Multiple pressure sensors are set up on the tire, and by establishing a position estimation model and a contact force estimation model, collecting and processing contact position and contact force data, the accurate detection of the force and contact position at the foot end is achieved.

Benefits of technology

The wheel leg robot realizes real-time perception of contact state in complex environments, improves travel efficiency and stability, provides reliable data support, and provides accurate data for navigation and obstacle avoidance tasks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118636196B_ABST
    Figure CN118636196B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of wheel-legged robots, and proposes a method and system for detecting the contact force and position of the foot of a wheel-legged robot. The method and system are implemented based on a plurality of pressure sensors arranged on a tire, and the method comprises the following steps: S1 collects the contact position data, pressure data group and pressure data of the tire, wherein the pressure data group is a collection of data detected by each pressure sensor at the same time; S2 calculates the characterization value corresponding to each pressure data group, and uses the characterization value and the contact position data as the first data set. The method and system for detecting the contact force and position of the foot of the wheel-legged robot, by arranging a plurality of pressure sensors on the tire to collect the force conditions of the robot's wheel legs during movement, and then outputs the contact force magnitude and contact position of the robot's foot according to the force conditions by two preset models, so as to realize accurate detection of the force and contact position at the foot.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of wheel-legged robots, and in particular to a method and system for detecting the contact force and position of the foot end of a wheel-legged robot. Background Art

[0002] Wheel-legged robots need to have real-time contact with the external environment when performing tasks. In unstructured terrain with strong randomness and high complexity, accurately sensing their contact status and obtaining contact position and contact force can help the robot identify obstacles in time and avoid collisions. By accurately sensing the contact conditions of the ground, obstacles, etc., the robot can flexibly adjust its behavior according to environmental changes and realize autonomous decision-making and path planning, which helps to improve the robot's travel efficiency and stability in different terrains.

[0003] At present, robots mainly use pressure sensors, force-sensitive resistors, strain sensors and other sensors to measure parameters such as contact force, contact area and contact time between them and the ground. These sensors are usually installed at the bottom of the wheels or the foot of the robot, and the contact state between the robot and the ground is determined by measuring the electrical signal output by the sensor. The existing detection method obtains information through the sensor embedded in the foot, and is mainly used for detection feedback on quadruped robots, but it cannot meet the requirements of foot shape detection, force detection and sensitivity of wheel-legged movement. Summary of the invention

[0004] In view of this, the present invention proposes a method and system for detecting the foot-end contact force and position of a wheel-legged robot. Multiple pressure sensors are set on the tires to collect the force conditions of the robot's wheel legs during movement. Two preset models then output the robot's foot-end contact force and contact position according to the force conditions, so as to achieve accurate detection of force and contact position at the foot.

[0005] The technical solution of the present invention is achieved in this way:

[0006] On the one hand, the present invention provides a method for detecting the contact force and position of the foot end of a wheel-legged robot, which is implemented based on a plurality of pressure sensors arranged on a tire, and the method comprises the following steps:

[0007] S1 collects tire contact position data, a pressure data group, and tire pressure data, wherein the pressure data group is a collection of data detected by each pressure sensor at the same time;

[0008] S2 calculates the characterization value corresponding to each pressure data group, takes the characterization value and the contact position data as a first data set, and takes the contact position data, the pressure data group and the tire pressure data as a second data set;

[0009] S3 establishes a first model and a second model, trains the first model with the first data set to obtain a position estimation model with the characterization value as input and the contact position estimation data as output, and trains the second model with the second data set to obtain a contact force estimation model with the contact position data and the pressure data set as input and the estimated pressure data of the tire as output;

[0010] S4 collects the detection data of each pressure sensor of the tire at the current moment to form real-time pressure data;

[0011] S5 converts the real-time pressure data into a characterization value and inputs it into a position estimation model to obtain current contact position estimation data;

[0012] S6 inputs the current contact position estimation data into the contact force estimation model to obtain the current estimated tire pressure data.

[0013] On the basis of the above technical solution, preferably, in a certain pressure data set, the characterization value λ is:

[0014]

[0015] Wherein, S1 and S2 represent the two largest pressure values ​​of the pressure data set respectively.

[0016] On the basis of the above technical solution, preferably, in step S3, the training of the first model by the first data set to obtain a position estimation model with the characterization value as input and the contact position estimation data as output includes the following steps:

[0017] S31 divides the first data set into a first training set and a first test set;

[0018] S32 trains the first model using the first training set to obtain an initial model for position estimation;

[0019] S33 evaluates the performance of the initial model for location estimation by cross-validation method;

[0020] S34 uses at least one indicator of mean square error and root mean square error to evaluate the performance of the initial location estimation model on the test set to obtain a location estimation model.

[0021] Further preferably, the basic equation of the first model is:

[0022]

[0023] Where λ is the characterization value, α is the position of the contact force on the tire, are polynomial coefficients, and d is the polynomial order.

[0024] Based on the above technical solution, preferably, in step S3, the second model is trained by the second data set to obtain a contact force estimation model with the contact position data and the pressure data group as input and the estimated pressure data of the tire as output, comprising the following steps:

[0025] S35: taking the pressure data group and the corresponding contact position data as samples, and adding corresponding tire pressure data labels to the samples to form a second data set;

[0026] S36 divides the second data set into a second training set, a second test set and a validation set;

[0027] S37 trains, tests and verifies the second model through the second training set, the second test set and the verification set respectively to obtain the contact force estimation model.

[0028] On the basis of the above technical solution, preferably, before step S1, by setting up an experimental platform, the rotation data of the tire is obtained to obtain the contact position data of the tire, the pressure sensor data inside the tire is collected to obtain a pressure data group, and the external force of the tire is collected to obtain the pressure data of the tire.

[0029] Further preferably, the experimental platform includes a contact plane arranged on a horizontal plane, a reciprocating pushing end that reciprocates toward the contact plane, a pressure collecting piece arranged on the contact plane, and a rotating detecting piece arranged on the reciprocating pushing end. The tire is arranged on the reciprocating pushing end and is continuously squeezed and leaves the contact plane by the reciprocating pushing end. During this process, the pressure data of the tire is obtained by the pressure collecting piece, and the rotating detecting piece obtains the relative rotation angle between the tire and the reciprocating pushing end.

[0030] On the basis of the above technical solution, preferably, the step S1 also includes removing abnormal values ​​and standardizing the collected tire contact position data, pressure data group and tire pressure data.

[0031] On the other hand, the present invention provides a foot contact force and position detection system for a wheel-legged robot, comprising a processor and a memory, wherein the processor is used to implement the above-mentioned foot contact force and position detection method for the wheel-legged robot when executing a computer program stored in the memory.

[0032] On the basis of the above technical solution, preferably, a plurality of strain gauge pressure sensors distributed in a circular shape with equal angles around the tire rotation axis are arranged in the tire.

[0033] The method and system for detecting the foot contact force and position of the wheel-legged robot of the present invention have the following beneficial effects compared with the prior art:

[0034] (1) Multiple pressure sensors are set on the tires to collect the force conditions of the robot's wheel legs during movement. The collected data are then input into the position estimation model to obtain the estimated position of the contact point. The estimated position of the contact point is then input into the contact force estimation model to obtain the estimated tire pressure, and then the robot's foot-end contact force and contact position are obtained, so as to achieve accurate detection of the force and contact position at the foot end, providing reliable data support for the walking and operation of the wheel-legged robot.

[0035] (2) By using the ratio of the difference between the deformation data of adjacent strain gauges and the larger one as the representation value, the data can be standardized, the influence of noise can be reduced, the feature expression ability can be enhanced, the model complexity can be simplified, and the generalization ability and model interpretability can be improved. At the same time, the representation value directly reflects the relative size of the deformation difference, making the position estimation model more stable and accurate during training, and having better versatility and easy interpretability, thereby improving the performance of the model in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 It is a schematic diagram of the steps of the method for detecting the foot end contact force and position of the wheel-legged robot of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] This method aims to achieve real-time perception of the ground conditions during the robot's movement by studying the position and force of the robot's foot contact, so as to ensure that it can move and operate accurately in complex environments. Through this ground contact detection method, it can ensure that the robot can stably contact the ground during movement to avoid unexpected situations, and provide accurate data support for the robot's navigation, path planning, obstacle avoidance and other tasks.

[0040] like Figure 1As shown, the method for detecting the foot contact force and position of the wheel-legged robot of the present invention is implemented based on a plurality of pressure sensors arranged on the tire, and specifically includes steps S1-S6.

[0041] Before performing step S1, it is necessary to build an experimental platform for collecting model training data, and through the set experimental platform, obtain the tire rotation data, collect the tire pressure sensor data, and obtain the tire pressure data. After collecting data through the experimental platform, subsequent model training can be carried out through the collected data to achieve real-time processing effects of the model during actual use, so that real-time wheel force position and force data can be obtained during the subsequent walking process of the wheel-legged robot.

[0042] Specifically, the experimental platform includes a contact plane arranged on a horizontal plane, a reciprocating pushing end that reciprocates toward the contact plane, a pressure collecting piece arranged on the contact plane, and a rotating detecting piece arranged on the reciprocating pushing end. The tire is arranged on the reciprocating pushing end and is continuously squeezed and leaves the contact plane by the reciprocating pushing end. During this process, the pressure data of the tire is obtained by the pressure collecting piece, and the rotating detecting piece obtains the relative rotation angle between the tire and the reciprocating pushing end.

[0043] When the experimental platform is built, a motor can be used to drive the crank to rotate, and a movable arm can be set on the cam. By limiting the movable direction of the end of the movable arm, a reciprocating push end is formed. The set plane can directly use a pad placed on the ground, and a pressure sensor is set on the pad as a pressure collection part. It is preferred to use a servo motor as the power source input, and drive the cam connecting rod mechanism after deceleration by the reducer to provide up and down reciprocating power to the robot's single leg to simulate the robot's motion contact with the ground. Every time the cam rotates one circle, the end of the robot's single leg realizes a contact with the ground, thereby realizing the simulation and research of contact perception. The data obtained is continuous and can more accurately reflect the changes in the contact state between the robot and the ground. By adjusting the speed of the motor and the deformation of the wheel bottom, it is also possible to simulate the impact effect of the robot on the road in different scenarios, which can improve the experimental accuracy of the robot.

[0044] For the rotating detection part, an encoder can be used. The encoder is set on the reciprocating push end. The encoder collects the relative rotation angle between the tire and the reciprocating push end, thereby determining the contact position of the tire and the contact plane, and realizing the data collection operation of the contact position.

[0045] It should be noted that during the experiment, since the tire contacts a fixed plane, the contact position obtained by the encoder is basically the actual contact position of the tire force. However, during the actual movement of the robot, the tire force is not always on a plane, so it is necessary to determine the tire force position by setting up multiple strain gauge pressure sensors.

[0046] Step S1: collecting tire contact position data, a pressure data group, and tire pressure data, wherein the pressure data group is a collection of data detected by each pressure sensor at the same time.

[0047] In this step, the collected tire contact position data is the tire rotation angle collected by the encoder on the experimental platform, and the tire contact position is determined by conversion. This data can only be obtained through the experimental platform, and cannot be obtained from the actual wheel-leg robot activity process. The collected pressure data group is the detection data of multiple pressure sensors arranged on the tire at the corresponding tire contact position and at the same time. The tire pressure data is detected by the pressure sensor arranged on the contact plane. Since the force is mutual, the force exerted on the fixed contact plane is equal to the force exerted on the tire. The tire force detection performed in this way has higher accuracy. In addition, the pressure sensor for detecting the overall force of the tire can also be set on the reciprocating pushing end, but considering that the tire has acceleration and it has a certain elasticity, there may be some external force influences, which leads to large errors in the collection of the overall tire force data.

[0048] In addition, this step also includes removing abnormal values ​​and standardizing the collected tire contact position data, pressure data group and tire pressure data.

[0049] Specifically, in order to ensure that the signals collected by the sensor are as free from external interference as possible, this solution sets the collection threshold to filter out invalid data. When the analog signal received by the host computer is too small, the acquired data is defined as invalid data and is reset to zero. At the same time, for the case where the sensor is not used continuously, the sensor is re-zeroed by subtracting the initial reading at each initialization. The standardization process is to eliminate the influence between different dimensions. Commonly used methods include Z-score standardization and Min-Max standardization.

[0050] Step S2: Calculate the characterization value corresponding to each pressure data set, use the characterization value and the contact position data as the first data set, and use the contact position data, the pressure data set and the tire pressure data as the second data set.

[0051] For the judgment of the contact position, in order to further explain in detail, taking a single point contact as an example, when the tire is compressed, among all the strain gauges, the strain gauge sensors on both sides of the contact point will produce the largest deformation. When the contact position is in the middle of the two strain gauges, the deformation of the two strain gauges is consistent. As the contact position is closer to the sensor positions at both ends, the strain difference between the two sensors will gradually increase. In a more special case, when the contact position is just on the strain gauge sensor, the deformation degree of this strain gauge sensor is the largest, and the deformation degree of the strain gauges on both sides of the corresponding contact point is equal. Therefore, by obtaining the strain gauge deformation data on both sides of the contact position, the foot end contact position can be accurately judged. In order to more clearly show this feature, taking two deformation gauges as an example, the ratio λ of the difference between the deformation data (s1, s2) of the adjacent strain gauges and the larger max(s1, s2) is taken as input, λ is the characterization value, and the contact position, that is, the angle α, is taken as the output, and the angle α represents the position of the contact force on the tire. Further fitting the relationship between them can be obtained:

[0052]

[0053] It should be noted that, in this embodiment, s1 and s2 are preferably the two largest strain gauge pressure values ​​in the pressure data set, but they do not have to be the two largest values. If two values ​​other than the largest are selected, there is a problem of reduced accuracy due to a large distance.

[0054] In the above formula, the range of the characterization value λ is within (-1,1). The closer the value is to the endpoint, the closer the contact point is to one of the two strain gauges. On the contrary, the closer the value is to 0, the closer the contact point is to the midpoint of the line between the two strain gauges.

[0055] Through multiple experiments on the experimental platform, the corresponding data of the characterization value λ and the angle α obtained when contacting at different positions are obtained, so that the next step can be carried out.

[0056] Step S3: Establish a first model and a second model, train the first model with the first data set to obtain a position estimation model with the characterization value as input and the contact position estimation data as output, train the second model with the second data set to obtain a contact force estimation model with the contact position data and pressure data group as input and the estimated pressure data of the tire as output.

[0057] According to multiple experiments on the experimental platform, the corresponding data of the characterization value λ and the angle α obtained when contacting at different positions are obtained, and a fitting model can be established. After the model is established, the contact position α can be more accurately estimated based on the data of the two strain gauge sensors. Among the many regression algorithms, the polynomial regression algorithm can model the nonlinear relationship between the input and output variables, and can obtain complex relationships that are difficult to model using linear regression. Compared with some other regression algorithms, its model establishment is also simpler and does not require a lot of computing power. It is also more suitable for robots that require fast real-time calculations. Therefore, in this embodiment, the first model is preferably modeled in a polynomial regression manner, and its basic equation is expressed as:

[0058]

[0059] Where λ is the characterization value, α is the position of the contact force on the tire, are polynomial coefficients, and d is the polynomial order.

[0060] In the actual experiment, when the same numerical normal phase load is applied, we obtain a set of λ and angle α. In order to show the relationship between λ and α, λ is used as input and α is used as output in all the obtained data. The order d of the polynomial regression is confirmed by comparing the fitting of different orders to the obtained data. Among them, the 5th order polynomial model has the best fit to the model and the lowest error, so the 5th order polynomial regression is preferably used for calculation in this embodiment.

[0061] For the second model, the SVR algorithm is preferred, which has stronger robustness for the estimation results of outliers and noise values, and has strong generalization ability. It is more suitable for robots that require high-dimensional input data and need to work in an environment with large external disturbances.

[0062] Specifically, step S3 of training the first model with the first data set to obtain a position estimation model with the characterization value as input and the contact position estimation data as output includes sub-steps S31-S34.

[0063] Step S31: Divide the first data set into a first training set and a first test set.

[0064] In this step, the first data set can be divided into a first training set and a first test set in a ratio of 8:2 for subsequent model training and testing.

[0065] Step S32: training the first model using the first training set to obtain an initial model for position estimation.

[0066] In this step, the first model is trained using the first training set, and the polynomial coefficients are estimated by a method such as the least squares method to obtain an initial model for position estimation.

[0067] Step S33: Evaluate the performance of the initial model for position estimation by a cross-validation method.

[0068] The cross-validation method is used to evaluate the performance of the initial location estimation model. The cross-validation method can use k-fold cross-validation to prevent overfitting problems.

[0069] Step S34: Use at least one indicator of mean square error and root mean square error to evaluate the performance of the initial location estimation model on the test set to obtain a location estimation model.

[0070] In this step, the performance of the model on the test set needs to be evaluated through indicators such as mean square error and root mean square error. After the evaluation, the model parameters with the best performance are saved to obtain the position estimation model.

[0071] In addition, the obtained position estimation model can be optimized through regularization and hyperparameter tuning.

[0072] Through the above steps, a polynomial regression model can be established and trained to estimate the contact position of the foot end of the wheel-legged robot. In the polynomial regression model, λ represents the ratio of the difference between the deformation data of adjacent strain gauge sensors to the larger one, which can effectively characterize the change of the contact position. In the process of model establishment, the characterization value λ is used as an input feature to help the algorithm better capture the relationship between the contact position and the sensor deformation data, thereby improving the prediction accuracy and stability of the model.

[0073] In step S3, the second model is trained by the second data set to obtain a contact force estimation model with the contact position data and the pressure data group as input and the estimated tire pressure data as output, including sub-steps S35-S37.

[0074] Step S35: taking the pressure data set and the corresponding contact position data as samples, and adding corresponding tire pressure data labels to the samples to form a second data set.

[0075] The pressure data set and the contact position data are used as samples, and the corresponding tire pressure data are added to the samples to form a second data set with complete feature-label pairs.

[0076] That is, the strain gauge sensor signal and the contact position estimation result are combined into a feature vector, and the pressure on the corresponding contact plane is added as a target label, and they are combined to form a complete second data set.

[0077] Step S36: Divide the second data set into a second training set, a second test set and a validation set.

[0078] The ratio of the second training set, the second test set and the validation set is preferably 70:15:15, which is the percentage in the second data set.

[0079] Step S37: training, testing and verifying the second model using the second training set, the second test set and the verification set respectively, to obtain a contact force estimation model.

[0080] The model is trained on the second training set to construct a preliminary contact force estimation model. The model performance is then evaluated on the validation set, and hyperparameter tuning and model optimization are performed. Finally, the model is evaluated on an independent test set to determine the actual performance and generalization ability of the model, thereby obtaining the final contact force estimation model.

[0081] The obtained contact force estimation model can estimate the contact force of the wheel during real-time operation, and thus provide reliable data for the robot's movements. For example, in robot grasping and manipulation tasks, the contact force can be calculated in real time through sensor data to ensure the safety and accuracy of the operation.

[0082] Step S4: Collect the detection data of each tire pressure sensor at the current moment to form real-time pressure data.

[0083] In this step, the wheel-legged robot walks in a real scene or a simulated scene, and the real-time pressure data obtained is real data, rather than simulated by the above-mentioned experimental platform. The real-time pressure data includes the data detected by multiple strain gauge sensors at the same time, which is equivalent to the pressure data group obtained in the experiment.

[0084] It should be noted that during the actual walking process of the wheel-legged robot, multiple strain gauge pressure sensors need to be installed in its tires for pressure monitoring.

[0085] Step S5: convert the real-time pressure data into a characterization value and input it into a position estimation model to obtain current contact position estimation data.

[0086] The real-time pressure data obtained in step S4 is selected and converted to obtain a corresponding characterization value, and the characterization value is input into the position estimation model obtained in step S3. The position estimation model outputs the current contact position estimation data. This data is an estimated value and may have a certain deviation from the true value, but this deviation can be basically ignored in actual operation. Even as the wheel-legged robot continues to move, the current contact position estimation data will continue to change and adjust.

[0087] It should be noted that there is no ideal plane in the actual walking process of the tire, so the contact point will not always be directly below the robot. In other words, the contact point position cannot be obtained by setting the encoder, and only the position of the point where the tire is closest to the ground can be obtained. In the previous experiment, since the tire moves back and forth on a plane, the point where the tire is close to the ground can be regarded as its contact point.

[0088] Step S6: Input the current contact position estimation data into the contact force estimation model to obtain the current estimated tire pressure data.

[0089] Unlike the model trained with experimental data, in this step, since the accurate contact point position cannot be obtained directly through the encoder, but the pressure must be obtained by the contact point position, the output of the position estimation model is directly used as the input of the contact force estimation model to obtain the estimated pressure data of the current tire.

[0090] The reason why the strain gauge sensor cannot obtain the force exerted on the tire based on the collected pressure is that the normal contact force will change with the change of the contact position (i.e., angle α). Therefore, it cannot be judged only by the deformation of the strain gauge. For example, when the same force is applied at different positions, the degree of deformation of the strain gauge is different. Therefore, the force at this time cannot be judged based on the strain gauge. The factor of the contact position α must also be considered comprehensively. Compared with the polynomial regression method, the detection of contact force needs to be able to better fit the nonlinear relationship and needs to be more suitable for high-dimensional input models. Therefore, the SVR algorithm is selected for modeling.

[0091] The foot contact force and position detection system of the wheel-legged robot of the present invention comprises a processor and a memory, wherein the processor is used to implement the above-mentioned foot contact force and position detection method of the wheel-legged robot when executing the computer program stored in the memory.

[0092] Specifically, the tire is provided with a plurality of strain gauge pressure sensors which are distributed in a circular shape with equal angles around the tire rotation axis.

[0093] In this embodiment, a collection system is set up. In order to obtain real-time contact information of the robot's foot, 12 strain gauge sensors are evenly arranged along the circumference inside the foot, that is, a strain gauge sensor is arranged every 30° inside the tire, and the inside of the tire is preferably between the inner tube and the outer tube of the tire.

[0094] After the tire touches the ground, the contact area will deform and the strain will be transmitted to the strain sensing element through the tire base. When the external pressure applied to the tire increases or decreases, the internal resistance of the sensor will also increase or decrease linearly and proportionally, resulting in the corresponding changing electrical signal being acquired by the acquisition system.

[0095] The acquisition system includes a controller, a Bluetooth remote communication module and a data acquisition conversion module. The Bluetooth module sends the collected signal to the robot host computer remotely by connecting the microcontroller board. The data acquisition conversion module is responsible for collecting strain signals and using two series-connected operation circuits to convert tiny voltage signals into a range that the controller can receive, and converts them into analog signals for output through ADC. In this way, the pressure data group and real-time pressure data in the above method can be obtained.

[0096] 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 in the protection scope of the present invention.

Claims

1. A method for detecting the contact force and position of the foot end of a wheel-legged robot, characterized in that: Based on a plurality of pressure sensors arranged on the tire, the method comprises the following steps: S1 collects tire contact position data, a pressure data group, and tire pressure data through an experimental platform, wherein the pressure data group is a collection of data detected by each pressure sensor at the same time; S2 calculates the characterization value corresponding to each pressure data group, takes the characterization value and the contact position data as the first data set, and takes the contact position data, the pressure data group and the tire pressure data as the second data set. In one of the pressure data groups, the characterization value λ is: ; Among them, S1 and S2 represent the two largest pressure values ​​of the pressure data set respectively; S3 establishes a first model and a second model, trains the first model with the first data set to obtain a position estimation model with the characterization value as input and the contact position estimation data as output, and trains the second model with the second data set to obtain a contact force estimation model with the contact position data and the pressure data set as input and the estimated pressure data of the tire as output; S4 collects the detection data of each pressure sensor of the tire at the current moment during the walking process of the wheel-legged robot to form real-time pressure data; S5 converts the real-time pressure data into a characterization value and inputs it into a position estimation model to obtain current contact position estimation data; S6 inputs the current contact position estimation data into the contact force estimation model to obtain the current estimated tire pressure data.

2. The method for detecting the foot end contact force and position of a wheel-legged robot as claimed in claim 1, characterized in that: In step S3, the first model is trained by the first data set to obtain a position estimation model with the characterization value as input and the contact position estimation data as output, including the following steps: S31 divides the first data set into a first training set and a first test set; S32 trains the first model using the first training set to obtain an initial model for position estimation; S33 evaluates the performance of the initial model for location estimation by cross-validation method; S34 uses at least one indicator of mean square error and root mean square error to evaluate the performance of the initial location estimation model on the test set to obtain a location estimation model.

3. The method for detecting the foot contact force and position of a wheel-legged robot as claimed in claim 2, characterized in that: The basic equation of the first model is: ; Where λ is the characterization value, α is the position of the contact force on the tire, are polynomial coefficients, and d is the polynomial order.

4. The method for detecting the foot end contact force and position of a wheel-legged robot according to claim 1, characterized in that: In step S3, the second model is trained by the second data set to obtain a contact force estimation model that takes the contact position data and the pressure data set as input and takes the estimated tire pressure data as output, including the following steps: S35: taking the pressure data group and the corresponding contact position data as samples, and adding corresponding tire pressure data labels to the samples to form a second data set; S36 divides the second data set into a second training set, a second test set and a validation set; S37 trains, tests and verifies the second model through the second training set, the second test set and the verification set respectively to obtain the contact force estimation model.

5. The method for detecting the foot end contact force and position of a wheel-legged robot as claimed in claim 1, characterized in that: Before step S1, an experimental platform is set up to obtain tire rotation data to obtain tire contact position data, collect tire internal pressure sensor data to obtain a pressure data group, and collect tire external force to obtain tire pressure data.

6. The method for detecting the foot end contact force and position of a wheel-legged robot as claimed in claim 5, characterized in that: The experimental platform includes a contact plane arranged on a horizontal plane, a reciprocating pushing end that reciprocates toward the contact plane, a pressure collecting piece arranged on the contact plane, and a rotating detecting piece arranged on the reciprocating pushing end. The tire is arranged on the reciprocating pushing end and is continuously squeezed and leaves the contact plane by the reciprocating pushing end. During this process, the pressure collecting piece obtains the pressure data of the tire, and the rotating detecting piece obtains the relative rotation angle between the tire and the reciprocating pushing end.

7. The method for detecting the foot end contact force and position of a wheel-legged robot as claimed in claim 1, characterized in that: The step S1 also includes removing abnormal values ​​and standardizing the collected tire contact position data, pressure data group and tire pressure data.

8. A foot contact force and position detection system for a wheel-legged robot, characterized in that: It comprises a processor and a memory, wherein the processor is used to implement the foot end contact force and position detection method of the wheel-legged robot as described in any one of claims 1-7 when executing the computer program stored in the memory.

9. The foot end contact force and position detection system of the wheel-legged robot as claimed in claim 8, characterized in that: The tire is provided with a plurality of strain gauge pressure sensors distributed in a circular manner with equal angles around the tire rotation axis.