A quadruped robot touch ground detection method, system, device and storage medium

Abstract

The application discloses a kind of four-legged robot touch ground detection method, system, equipment and storable medium, by designing the disturbance observer of generalized momentum, four-legged robot can estimate itself generalized momentum in real time, acutely perceive external environment change, especially the momentum change caused by external force, this technical solution makes robot without additional sensor can accurately judge leg touch ground condition, improves the adaptability to complex terrain, whether flat or rugged terrain, can according to motion state and external environment adjustment walking strategy, keep stability, by comparing actual and expected trajectory difference, robot can adjust walking posture and pace in time, ensure walking smoothly and efficiently, quickly and accurately reach destination, in addition, accurately judge touch ground reduces the risk of falling or damage, reduces maintenance cost and use risk.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot ground contact detection, and in particular to a quadruped robot ground contact detection method, system, device and storage medium. BACKGROUND

[0002] As a product of advanced technology that imitates animal walking, the quadruped robot has shown great application potential in disaster relief, exploration, logistics distribution and other fields. However, in order to effectively walk and work in these complex environments, the quadruped robot needs to have accurate ground sensing capability, especially the accuracy of foot end ground contact detection, which is directly related to the working efficiency and safety of the robot.

[0003] Traditional ground contact detection methods, such as hydraulic quadruped robot ground contact detection, usually rely on direct detection by pressure sensors or contact sensors, or indirect detection by force sensors and displacement sensors. Although these methods are effective, they may encounter problems such as sensor wiring difficulty and limited detection accuracy in some complex terrains or step terrains. SUMMARY

[0004] Therefore, the present application provides a quadruped robot ground contact detection method, system, device and storage medium, which can effectively solve the defects of sensor wiring difficulty and limited detection accuracy in the prior art.

[0005] The technical solution of the present application is as follows:

[0006] A quadruped robot ground contact detection method, comprising:

[0007] establishing a quadruped robot dynamics model;

[0008] designing a generalized momentum disturbance observer based on the quadruped robot dynamics model, the generalized momentum disturbance observer being capable of estimating the generalized momentum of the quadruped robot in real time and detecting the change in the generalized momentum caused by external forces;

[0009] obtaining actual torque values and leg position information at leg joints of the quadruped robot;

[0010] calculating the expected trajectory of the leg under a given control input according to the disturbance observer and the dynamics model of the quadruped robot;

[0011] using a discrete-time integral formula to convert the actual observed joint torque and leg position information into an actual trajectory, and comparing the difference between the actual trajectory and the expected trajectory;

[0012] detecting whether the leg has contacted the ground according to the difference result.

[0013] As a further optional solution of the four-legged robot ground contact detection method, the actual observed joint torque and leg position information are converted into an actual trajectory using a discrete-time integration formula, specifically including:

[0014] Obtaining discrete-time points of observed joint torque and leg position information;

[0015] For each discrete-time point, using a discrete-time integration formula to calculate the state update of the time point, accumulating the state update of each discrete-time point;

[0016] Connecting the state updates of the discrete-time points to form a continuous actual trajectory.

[0017] As a further optional solution of the four-legged robot ground contact detection method, the method further includes using a Kalman filter to fuse the leg state expectation based on the gait scheduler with the probability of the ground height model.

[0018] As a further optional solution of the four-legged robot ground contact detection method, the ground height model models the average foot vertical force at the start of contact according to a Gaussian normal distribution, and infers the probability of foot contact with the ground according to the model.

[0019] As a further optional solution of the four-legged robot ground contact detection method, the ground height model models the average foot vertical force at the start of contact according to a Gaussian normal distribution, and infers the probability of foot contact with the ground according to the model, specifically including:

[0020] Collecting vertical force data when the foot is in contact with the ground;

[0021] Calculating the average foot vertical force at the start of each contact by calculating the collected data;

[0022] Calculating the mean and standard deviation of the average foot vertical force;

[0023] Using the mean and standard deviation to construct a Gaussian distribution function;

[0024] Determining a contact threshold;

[0025] Using the Gaussian distribution function to calculate the probability that the foot vertical force is greater than or equal to the contact threshold;

[0026] According to the calculated contact probability, the probability of foot contact with the ground can be inferred.

[0027] As a further optional solution of the four-legged robot ground contact detection method, the method further includes:

[0028] According to the result of ground contact detection, modifying the gait control action of the four-legged robot;

[0029] When the leg is detected to touch the ground, the control strategy of the leg is adjusted according to the time and position of the ground touch.

[0030] A quadruped robot ground touch detection system comprises:

[0031] A dynamics model establishment module is configured to establish a dynamics model describing dynamic behaviors of the quadruped robot in various motion states.

[0032] A generalized momentum disturbance observer module is designed based on the dynamics model and is capable of estimating the generalized momentum of the quadruped robot in real time and detecting changes in the generalized momentum caused by external forces.

[0033] A data acquisition module is configured to acquire actual torque values at joints of legs of the quadruped robot and leg position information.

[0034] An expected trajectory calculation module is configured to calculate expected trajectory results of the legs under given control inputs according to the generalized momentum disturbance observer and the dynamics model.

[0035] An actual trajectory generation and comparison module is configured to convert the actually observed joint torque and leg position information into actual trajectories using a discrete-time integral formula and compare the actual trajectories with the expected trajectories to detect whether the legs touch the ground.

[0036] A ground touch detection result output module is configured to output results of the ground touch detection.

[0037] As a further optional solution of the quadruped robot ground touch detection system, the system further comprises:

[0038] A gait control action modification module is configured to modify gait control actions of the quadruped robot according to results of the ground touch detection.

[0039] A leg control strategy adjustment module is configured to adjust the control strategy of the leg according to the time and position of the ground touch when the leg is detected to touch the ground.

[0040] A computing device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements steps of the quadruped robot ground touch detection method of any one of the above when executing the computer program.

[0041] A computer readable storage medium stores a computer program, and the computer program implements steps of the quadruped robot ground touch detection method of any one of the above when executed by a processor.

[0042] The beneficial effects of the present application are: by designing a generalized momentum disturbance observer, the generalized momentum of the quadruped robot can be estimated in real time, which helps the robot to more accurately understand its own motion state, the disturbance observer can not only estimate the generalized momentum, but also detect the change of the generalized momentum caused by external force (such as ground reaction force), which enables the robot to more sensitively perceive the change of the external environment, so as to more accurately judge whether the leg touches the ground without sensors, since the technical solution can estimate the generalized momentum in real time and detect the change of the external force, the robot can better adapt to complex terrain, whether it is flat ground or rugged mountain road, the robot can adjust the walking strategy according to its own motion state and the change of the external environment, so as to maintain stability, by comparing the difference between the actual trajectory and the expected trajectory, the robot can timely adjust the walking posture and pace to ensure the smooth and efficient walking, which helps the robot to be more rapid and accurate in reaching the destination when performing tasks, since the robot can more accurately judge whether the leg touches the ground, it can reduce the occurrence of falling or damage caused by misjudgment, which helps to reduce the maintenance cost and use risk of the robot. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0044] Fig. 1 The flowchart of the method for detecting the ground contact of the quadruped robot according to the present application;

[0045] Fig. 2 The composition diagram of the system for detecting the ground contact of the quadruped robot according to the present application;

[0046] Fig. 3 The composition diagram of the computing device according to the present application. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0048] REFERENCE Figs. 1 to 3 A method for detecting the ground contact of a quadruped robot, comprising:

[0049] A dynamics model of the quadruped robot is established;

[0050] Based on the dynamics model of the quadruped robot, a disturbance observer of generalized momentum is designed, which can estimate the generalized momentum of the quadruped robot in real time and detect the change of the generalized momentum caused by external force;

[0051] Actual torque values at joints of legs of the quadruped robot and leg position information are obtained;

[0052] According to the disturbance observer and the dynamics model of the quadruped robot, an expected trajectory result of the leg under a given control input is calculated;

[0053] The actually observed joint torque and leg position information are converted into an actual trajectory by using a discrete-time integral formula, and the difference between the actual trajectory and the expected trajectory is compared;

[0054] Whether the leg touches the ground is detected according to the difference result.

[0055] In this embodiment, the disturbance observer of generalized momentum is designed, which can estimate the generalized momentum of the quadruped robot in real time, which helps the robot to more accurately understand its own motion state. The disturbance observer can not only estimate the generalized momentum, but also detect the change of the generalized momentum caused by external force (such as ground reaction force), which enables the robot to more sensitively perceive the change of the external environment, so as to more accurately judge whether the leg touches the ground without sensors. Since the technical solution can estimate the generalized momentum in real time and detect the change of the external force, the robot can better adapt to complex terrain, whether it is flat ground or rugged mountain road. The robot can adjust the walking strategy according to its own motion state and the change of the external environment, so as to maintain stability. By comparing the difference between the actual trajectory and the expected trajectory, the robot can timely adjust the walking posture and pace to ensure smooth and efficient walking, which helps the robot to more quickly and accurately reach the destination when performing tasks. Since the robot can more accurately judge whether the leg touches the ground, it can reduce the occurrence of falling or damage caused by misjudgment, which helps to reduce the maintenance cost and use risk of the robot.

[0056] Preferably, the actual observed joint torque and leg position information are converted into an actual trajectory by using a discrete-time integral formula, specifically including:

[0057] Discrete-time points of the observed joint torque and leg position information are obtained;

[0058] For each discrete-time point, a state update of the time point is calculated by using a discrete-time integral formula, and state updates of each discrete-time point are accumulated;

[0059] The state updates at discrete time points are connected to form a continuous actual trajectory.

[0060] In this embodiment, by obtaining the discrete time points of observed joint torque and leg position information, the data can be ensured to be representative in time, which are selected according to the sampling frequency of the system or specific requirements, so as to capture the key motion information, for each discrete time point, the state update at this time point is calculated using the discrete time integral formula, which can accurately calculate the state at the current time point based on the state information at the previous time point and the observation data at the current time point, by accumulating the state update at each discrete time point, the state change in the whole motion process can be gradually constructed, and the state updates at discrete time points can be connected to form a continuous actual trajectory, which not only reflects the motion state of the robot at each discrete time point, but also makes the trajectory more continuous and smooth in time through interpolation or smoothing processing, using the discrete time integral formula for state update can realize real-time processing of observation data, which means that the robot can immediately calculate and update after receiving new observation data, so as to realize real-time tracking and monitoring of motion state, the discrete time integral formula is usually easier to implement and calculate than the continuous time integral formula, because it only needs to consider a limited number of discrete time points, without integrating over the entire time domain. This greatly reduces the amount of calculation and improves the calculation efficiency, the discrete time integral formula has certain anti-interference ability to noise and interference in observation data, because it only considers the data at discrete time points, so it can filter out some high-frequency noise and random interference, thereby improving the reliability and stability of the trajectory; The technical scheme can adapt to different types of observation data and motion state, whether it is joint torque or leg position information, as long as it can be processed by the discrete time integral formula, it can be used to construct the actual trajectory, which makes the technical scheme have strong adaptability and flexibility.

[0061] It should be noted that the specific steps are:

[0062] Selecting integral method: according to the characteristics and requirements of the data, selecting appropriate discrete time integral method, including forward Euler method, backward Euler method and trapezoidal method, etc.

[0063] Apply integral formula: for each time point, use the discrete time integral formula to calculate the state update at this time point, for example, when using the forward Euler method, the state update formula is: y(k) = y(k-1) + t*u(k-1), where y(k) represents the state at the current time point, y(k-1) represents the state at the previous time point, t represents the sampling period, u(k-1) represents the input at the previous time point (here is some form of joint torque or leg position information).

[0064] Cumulative state update: By continuously applying the integral formula, the state update at each time point is accumulated, thereby obtaining the actual trajectory of the entire time series;

[0065] Trajectory reconstruction: Connecting the state updates at discrete time points to form a continuous actual trajectory. It may be necessary to smooth the trajectory to eliminate fluctuations or jumps caused by discrete integration.

[0066] Preferably, the method further includes using a Kalman filter to fuse the leg state predictions based on the gait scheduler with the probabilities of the ground height model.

[0067] In this embodiment, the Kalman filter can fuse the leg state prediction from the gait scheduler and the probabilistic information from the ground height model. This fusion strategy utilizes data from two different sources, making the prediction results more comprehensive and accurate. The gait scheduler provides the leg state prediction based on robot motion planning and gait strategy, while the ground height model provides the probability distribution of ground height based on environmental perception and terrain information. Through the iterative update process of the Kalman filter, the prediction of the leg state can be continuously optimized. The filter continuously adjusts the weights and uncertainties of the prediction results based on new observation data and prior information, thereby obtaining a more accurate leg state estimate. After introducing the Kalman filter, the robot can better adapt to complex terrain. Whether on flat ground or rugged mountain roads, the robot can flexibly adjust its walking strategy based on its own motion state, the gait scheduler's expectations, and the probability information of the ground height model. Through accurate leg state prediction and terrain perception, the robot can plan its walking path and gait strategy more effectively, which helps reduce unnecessary energy consumption and wear, and improve walking efficiency and stability. Prediction results that fuse multiple information sources are generally more reliable than prediction results from a single source. The Kalman filter improves the reliability of leg state prediction by fusing the gait scheduler's expectations and the probability information of the ground height model.

[0068] It should be noted that the specific steps are as follows:

[0069] Define system state: System state x can include parameters such as leg position and velocity. These parameters are defined based on the expectations of the gait scheduler. For example, state x can be a vector containing the position and velocity information of the leg in three-dimensional space.

[0070] Define observation: Observation z comes from the probability distribution of the ground height model, which can be represented as the height information when the leg actually contacts the ground. Observation z may contain noise, so a probability distribution is needed to describe its uncertainty.

[0071] Initializing Kalman filter parameters: including state transition matrix F, which describes how the system state transitions from the previous time step to the current time step, observation matrix H, which describes how the system state maps to the observation space, process noise variance matrix Q, which describes the uncertainty in the state transition process, observation noise variance matrix R, which describes the uncertainty in the observation process, initial state vector x0, and initial error covariance matrix P0, which are set according to the initial conditions of the gait scheduler;

[0072] Prediction step: using the state transition matrix F and the state estimate at the previous time step x_(k-1) to predict the state at the current time step x_k_pred, updating the error covariance matrix P_k_pred to reflect the uncertainty of the prediction;

[0073] Update step: calculate the Kalman gain K, which represents the degree of influence of observation on state estimation, use the observation matrix H, observation z, and predicted error covariance matrix P_k_pred to calculate K, update the state estimate x_k, and take the weighted sum of the predicted value x_k_pred and the observed value z as the new state estimate, update the error covariance matrix P_k to reflect the uncertainty of the updated state estimate;

[0074] Iterative Kalman filtering process: for each time step, repeat the prediction and update steps, and over time, the Kalman filter will gradually converge to the true state of the system;

[0075] Fusing the probability of the ground height model: in the update step, the observation z comes from the probability distribution of the ground height model, and the Kalman filter fuses the probability of the ground height model with the expectation of the gait scheduler by calculating the Kalman gain K, and the fusion result is reflected in the updated state estimate x_k and error covariance matrix P_k.

[0076] Preferably, the ground height model models the average foot vertical force at the start of contact according to a Gaussian normal distribution, and infers the probability of foot-ground contact according to this model.

[0077] Preferably, the ground height model models the average foot vertical force at the start of contact according to a Gaussian normal distribution, and infers the probability of foot-ground contact according to this model, specifically including:

[0078] Collecting vertical force data when the foot is in contact with the ground;

[0079] Calculating the collected data to calculate the average foot vertical force at the start of each contact;

[0080] calculate the mean and standard deviation of the average foot vertical force;

[0081] construct a Gaussian distribution function using the mean and standard deviation;

[0082] determine the contact threshold;

[0083] calculate the probability of the foot vertical force being greater than or equal to the contact threshold using the Gaussian distribution function;

[0084] According to the calculated contact probability, the probability of the foot contacting the ground can be inferred.

[0085] In this embodiment, when the foot contacts the ground, the vertical force data received by the foot is collected in real time, and the vertical force data collected at the start of each contact is calculated to obtain the average foot vertical force, which helps to eliminate the randomness and uncertainty of single measurement; the average foot vertical force of multiple contacts is statistically analyzed, and the mean and standard deviation are calculated, which are the core parameters of the Gaussian distribution function, determining the shape and position of the distribution, and using the calculated mean and standard deviation, a Gaussian distribution function is constructed, which can describe the probability distribution of the foot vertical force; according to the actual needs, a reasonable contact threshold is determined, which is used to judge whether the foot contacts the ground, and using the Gaussian distribution function, the probability of the foot vertical force being greater than or equal to the contact threshold is calculated, which reflects the possibility of the foot contacting the ground, and according to the calculated contact probability, combined with other related information (such as the expected gait scheduler, the ground height model, etc.), the probability of the foot contacting the ground is comprehensively inferred.

[0086] It should be noted that through Gaussian distribution modeling and probability calculation, the foot can be more accurately judged whether it contacts the ground, and this method considers the randomness and uncertainty of the vertical force, improving the accuracy of detection; the ground height model combines the Gaussian distribution function, so that the robot can better adapt to different terrains, whether it is flat ground or rugged mountain road, the robot can judge and adjust according to the distribution of the foot vertical force; accurate contact detection helps the robot to maintain a stable walking state, and during walking, the robot can adjust the gait and posture in time according to the contact between the foot and the ground, to avoid falling or damage; through accurate contact detection, the robot can more reasonably allocate energy and plan the walking path, which helps to reduce energy consumption and improve walking efficiency.

[0087] Preferably, the method further comprises:

[0088] According to the result of the touch detection, the gait control action of the quadruped robot is modified;

[0089] When the leg is detected to touch the ground, the control strategy of the leg is adjusted according to the time and position of the touch.

[0090] In this embodiment, through real-time ground contact detection, the robot can accurately determine when and where each leg touches the ground. This enables the robot to adjust the gait and posture in a timely manner based on the ground contact information, thereby maintaining a more stable walking state; based on the results of ground contact detection, the robot can optimize its gait control actions, for example, if it detects that a certain leg touches the ground prematurely, the robot can adjust the pace or force of the other legs to maintain the coordination and balance of the overall gait, which helps to reduce energy consumption during walking and improve walking efficiency; the robot can better adapt to different terrains based on the information from ground contact detection, for rough and uneven ground, the robot can adjust the control strategy of the legs to avoid falling or damage, in addition, for different materials and hardness of the ground, the robot can also adjust the contact force and gait mode of the feet to maintain the best walking effect; during walking, the robot may encounter various obstacles, through ground contact detection, the robot can perceive the existence of obstacles in advance and take corresponding obstacle avoidance measures, which helps to reduce collisions with obstacles and protect the safety and integrity of the robot; this technical solution enables the robot to make intelligent decisions based on real-time ground contact detection information, this intelligence not only reflects in gait control and obstacle avoidance ability, but also reflects in the perception and adaptation to the environment, the robot can continuously learn and adjust its walking strategy to cope with different challenges and scenarios; for users using quadruped robots, stable walking and efficient gait control are one of the key factors to improve user experience, by implementing this technical solution, the robot can walk more intelligently and stably, thereby providing better user experience.

[0091] A quadruped robot ground contact detection system, comprising:

[0092] A dynamics model establishment module for establishing a dynamics model describing the dynamic behavior of the quadruped robot in various motion states;

[0093] A generalized momentum disturbance observer module designed based on the dynamics model, capable of estimating the generalized momentum of the quadruped robot in real time and detecting changes in the generalized momentum caused by external forces;

[0094] A data acquisition module for acquiring actual torque values and leg position information at the joints of the quadruped robot;

[0095] An expected trajectory calculation module for calculating the expected trajectory of the legs under given control input based on the generalized momentum disturbance observer and the dynamics model;

[0096] An actual trajectory generation and comparison module for converting the actual observed joint torque and leg position information into actual trajectories using a discrete-time integral formula, and comparing the actual trajectories with the expected trajectories to detect whether the legs have touched the ground;

[0097] a ground contact detection result output module configured to output a result of the ground contact detection.

[0098] Preferably, the system further comprises:

[0099] a gait control action modification module configured to modify a gait control action of the quadruped robot according to the result of the ground contact detection;

[0100] a leg control strategy adjustment module configured to adjust a control strategy of a leg according to a time and a position of the ground contact when the ground contact of the leg is detected.

[0101] A computing device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the ground contact detection method of any one of the above quadruped robots when executing the computer program.

[0102] A computer readable storage medium, wherein the storage medium stores a computer program, and the computer program implements the steps of the ground contact detection method of any one of the above quadruped robots when executed by a processor.

[0103] The above merely describes preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A four-legged robot touch-down detection method, characterized by, The method comprises: establishing a quadruped robot dynamics model; designing a generalized momentum disturbance observer based on the quadruped robot dynamics model, which can estimate the generalized momentum of the quadruped robot in real time and detect changes in the generalized momentum caused by external forces; obtaining actual torque values at the joints of the legs of the quadruped robot and leg position information; calculating the expected trajectory of the legs under a given control input according to the disturbance observer and the dynamics model of the quadruped robot; using a discrete-time integral formula to convert the actual observed joint torque and leg position information into an actual trajectory, and comparing the difference between the actual trajectory and the expected trajectory; detecting whether the leg has touched the ground according to the difference result; wherein the actual observed joint torque and leg position information are converted into an actual trajectory using a discrete-time integral formula, which specifically comprises: obtaining discrete-time points of observed joint torque and leg position information; for each discrete-time point, using a discrete-time integral formula to calculate the state update at that time point, and accumulating the state updates at each discrete-time point; connecting the state updates at the discrete-time points to form a continuous actual trajectory; the method further comprises using a Kalman filter to fuse the leg state expectation based on the gait scheduler with the probability of the ground height model; the ground height model models the average foot vertical force at the start of contact according to a Gaussian normal distribution, and infers the probability of foot contact with the ground according to the model; the ground height model models the average foot vertical force at the start of contact according to a Gaussian normal distribution, and infers the probability of foot contact with the ground according to the model, which specifically comprises: collecting vertical force data when the foot is in contact with the ground; calculating the average foot vertical force at the start of each contact by calculating the collected data; calculating the mean and standard deviation of the average foot vertical force; constructing a Gaussian distribution function using the mean and standard deviation; determining a contact threshold; using the Gaussian distribution function to calculate the probability that the foot vertical force is greater than or equal to the contact threshold; according to the calculated contact probability, the probability of foot contact with the ground can be inferred. 2.The touch-down detection method of the quadruped robot according to claim 1, wherein, The method further comprises: modifying the gait control action of the quadruped robot according to the result of the ground contact detection; when the leg is detected to touch the ground, adjusting the control strategy of the leg according to the time and position of the ground contact.

3. A quadruped robot touch-down detection system, comprising: The method comprises: a dynamics model establishment module for establishing a dynamics model describing the dynamic behavior of the quadruped robot in various motion states; a generalized momentum disturbance observer module designed based on the dynamics model, which can estimate the generalized momentum of the quadruped robot in real time and detect changes in the generalized momentum caused by external forces; a data acquisition module for obtaining actual torque values at the joints of the legs of the quadruped robot and leg position information; an expected trajectory calculation module for calculating the expected trajectory of the legs under a given control input according to the generalized momentum disturbance observer and the dynamics model; an actual trajectory generation and comparison module for converting the actual observed joint torque and leg position information into an actual trajectory using a discrete-time integral formula, and comparing the actual trajectory with the expected trajectory to detect whether the leg has touched the ground; The touch-down detection result output module is configured to output a result of the touch-down detection. The actual trajectory is converted from the observed joint torque and leg position information using a discrete-time integral formula, and specifically includes: Discrete-time points of the observed joint torque and leg position information are obtained; For each discrete-time point, a state update at the time point is calculated using a discrete-time integral formula, and state updates at each discrete-time point are accumulated; The state updates at the discrete-time points are connected to form a continuous actual trajectory; The system further includes using a Kalman filter to fuse the leg state expectation based on the gait scheduler with the probability of the ground height model; The ground height model models the average foot vertical force at the start of contact according to a Gaussian normal distribution, and infers the probability of the foot contacting the ground according to the model; The ground height model models the average foot vertical force at the start of contact according to a Gaussian normal distribution, and infers the probability of the foot contacting the ground according to the model, and specifically includes: Vertical force data is collected when the foot contacts the ground; The collected data is calculated to calculate the average foot vertical force at the start of each contact; The mean and standard deviation of the average foot vertical force are calculated; The Gaussian distribution function is constructed using the mean and standard deviation; A contact threshold is determined; The probability of the foot vertical force being greater than or equal to the contact threshold is calculated using the Gaussian distribution function; According to the calculated contact probability, the probability of the foot contacting the ground can be inferred.

4. The touch-down detection system of a quadruped robot according to claim 3, wherein, The system further includes: A gait control action modification module configured to modify the gait control action of the quadruped robot according to the result of the touch-down detection; A leg control strategy adjustment module configured to adjust the control strategy of the leg according to the time and position of the touch-down when the touch-down of the leg is detected.

5. A computing device, comprising: A computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the quadruped robot touch-down detection method of any one of claims 1-2.

6. A computer-readable storage medium, characterized in that, The storage medium has a computer program stored thereon, and the computer program is executed by the processor to implement the steps of the quadruped robot touch-down detection method of any one of claims 1-2.

Images