Hexapod robot terrain adaptive walking control method and system based on multi-sensor fusion

Abstract

The invention provides a terrain self-adaptive walking control method for a hexapod robot based on multi-sensor fusion, which is characterized in that an inertial measurement unit IMU1 and an inertial measurement unit IMU2 are respectively arranged at the sole and the handle of each foot of the hexapod robot, and the inertial measurement units are used for measuring the acceleration and the angular acceleration of corresponding parts; the control method specifically comprises the following steps that 1, IMU1 and IMU2 on each foot are used for collecting data, and a foot end attitude matrix of each foot is obtained; step 2, fusing the foot end attitude matrix into the current state of the hexapod robot to form an updated body state; 3, the updated body state information is transmitted into a quadratic programming solver, an optimal ground reaction force distribution matrix F is obtained, the reaction force of each foot is obtained, a joint acting force Fj is calculated according to the obtained ground reaction force distribution matrix F, a corresponding joint control instruction is converted according to the joint acting force Fj, and the joint control instruction is calculated according to the joint acting force Fj. And then a joint control command obtained through calculation is sent to a joint motor of the hexapod robot to control actions of legs of the hexapod robot, and the state of a body of the hexapod robot is kept stable.

Description

Technical Field

[0001] The present invention relates to the field of robot control, and in particular to a terrain adaptive walking control method and system for a hexapod robot based on multi-sensor fusion. Background Art

[0002] With the rapid development of robotics, hexapod robots have garnered widespread attention due to their exceptional adaptability and stability in complex terrain. These robots are capable of navigating uneven and rugged terrain that wheeled or tracked systems struggle to handle, demonstrating their enormous potential for exploration, rescue, and military applications. Compared to traditional mobile platforms, hexapod robots offer enhanced maneuverability and are capable of maintaining stability in extreme conditions, such as rocky terrain, muddy terrain, and slopes. Furthermore, their multi-point contact with the ground allows hexapod robots to maintain overall balance even if a single supporting leg loses support, significantly improving the system's robustness and reliability.

[0003] However, despite the many advantages of hexapod robots, they still face several challenges in practical applications, especially in terms of perceiving the environment and maintaining a stable gait. First of all, in complex natural environments, accurate perception of the terrain is crucial. Traditional solutions usually rely on external sensors such as laser radar (LiDAR), visual cameras, etc. to build a three-dimensional model of the surrounding environment. Although this method is effective, it has obvious limitations. For example, in the presence of smoke, dust, vegetation or insufficient light, the information provided by external sensors becomes unreliable or even invalid. In addition, even under ideal conditions, high-precision 3D modeling often requires a lot of computing resources and time, which places higher demands on real-time control. Therefore, how to achieve efficient and stable walking in the absence of reliable external perception has become an urgent problem to be solved.

[0004] To overcome these challenges and enhance the adaptability of hexapod robots in unknown or dynamically changing environments, researchers have begun exploring methods that combine advanced sensing technologies with intelligent control algorithms. Advances in sensing technology enable hexapod robots to acquire richer and more accurate ground information by integrating a variety of sensors. For example, the use of diverse sensing devices such as strain gauge force sensors, sensitive skin, optical force sensors, and tentacles can monitor the pressure distribution, tilt angle, and surface characteristics of each leg in real time when it contacts the ground, providing critical data for the control system. These sensorized foot-end designs not only enhance the robot's terrain recognition capabilities but also allow the controller to adjust according to actual contact conditions, thereby optimizing gait planning and posture control.

[0005] In this context, Virtual Model Control (VMC) has been widely studied as an effective control framework. The core idea of ​​VMC is to use a simplified dynamic model to simulate the motion behavior under ideal conditions and adjust the actual robot's movements through a feedback mechanism to keep it as close to this ideal state as possible. This simplified model has limitations when dealing with complex, uneven, or dynamically changing terrain. However, VMC's advantage lies in its ability to quickly implement control strategies and provide real-time state feedback based on sensor data. However, it is sensitive to noise and errors in sensor data, which can affect control effectiveness.

[0006] Therefore, how to develop a new control method that can effectively integrate multiple sensor data and adapt to complex and changing environments has become a key issue that needs to be urgently solved in the current field of hexapod robot research. Summary of the Invention

[0007] The main purpose of the present invention is to provide a terrain adaptive walking control method and system for a hexapod robot based on multi-sensor fusion, which is used to meet the control requirements of the hexapod robot.

[0008] To achieve the above objectives, the present invention adopts a technical solution: a terrain-adaptive walking control method for a hexapod robot based on multi-sensor fusion, wherein an inertial measurement unit (IMU1) and an inertial measurement unit (IMU2) are respectively provided on the sole and handle of each foot of the hexapod robot. The inertial measurement units are used to measure the acceleration and angular acceleration of the corresponding parts. The method specifically comprises the following steps:

[0009] Step 1: Use IMU1 and IMU2 on each foot to collect data and obtain the foot-end posture matrix of each foot;

[0010] Step 2: The foot-end posture matrix is ​​integrated into the current state of the hexapod robot to form an updated body state;

[0011] Step 3: Pass the updated body state information into the quadratic programming solver to obtain the optimal ground reaction force distribution matrix F, and then calculate the joint force F according to the obtained ground reaction force distribution matrix F. j , and then send the calculated joint control commands to the joint motors of the hexapod robot to control the movement of the hexapod robot legs and keep the hexapod robot body stable.

[0012] Preferably, step one specifically includes the following steps:

[0013] The step 1 specifically includes the following steps:

[0014] Step 11. Use a low-pass filter to estimate the bias of IMU1 and IMU2 on the same leg, as shown in the following formula:

[0015]

[0016] where Δω k+1 is the bias of IMU1 and IMU2 at the next moment; Δω k The bias of IMU1 and IMU2 at the current moment, β is the coefficient of the low-pass filter, which is used to balance the weight of the bias of IMU1 and IMU2 at the previous moment and the bias of IMU1 and IMU2 at the current moment; is the angular velocity measurement value at the current moment;

[0017] Step 12: Determine whether the hexapod robot's feet are in a stationary state:

[0018] If the acceleration in the direction of gravity measured by the accelerometers in IMU1 and IMU2 is close to the acceleration due to gravity, that is, |ag<0.1, where a is the measured acceleration and g is the acceleration due to gravity, it indicates that the legs of the hexapod robot are currently stationary. Spherical Linear Interpolation (SLERP) is used to correct the acceleration, as shown in the following formula:

[0019] Δq corr =SLERP(q I ,Δq acc ; α a ) (2),

[0020] where Δq corr is the corrected quaternion, Δq acc is the quaternion measured by the accelerometers in IMU1 and IMU2, which represents the difference between the gravity direction measured by the accelerometer and the current posture; I is the unit quaternion ([1,0,0,0] T ), indicating a state without any rotation; α a is the interpolation coefficient, and the current posture quaternion Δq is obtained corr ;

[0021] If the accelerometers in IMU1 and IMU2 measure an acceleration in the direction of gravity in the world coordinate system close to the acceleration of gravity, that is, |ag|>0.1, it indicates that the hexapod robot's feet are currently in a swinging state. In this case, the current angular velocity is calculated directly using the bias of IMU1 and IMU2 estimated by the low-pass filter, as shown in the following formula:

[0022]

[0023] where ω corr is the corrected angular velocity, is the current angular velocity measured by the gyroscope, Δω k+1 is the bias of IMU1 and IMU2 at the next moment, and then through ω corr The current attitude angles φ and θ of IMU1 and IMU2 are obtained by integration, and then the attitude quaternion Δq is obtained by Euler angle quaternion transformation. corr ;

[0024] Step 13: Calculate the drift of the foot end relative to the world coordinate system used by the robot. The expression is as follows:

[0025]

[0026] where Δq W′W is the quaternion of the world coordinate offset, Δq corr is the corrected drift quaternion, and the two are multiplied to obtain the corrected drift quaternion Δq W′ ;

[0027] Step 14: Calculate the ankle joint motion as follows

[0028]

[0029] Among them, q FA is the posture of the foot relative to the ankle joint; q AS is the posture of the ankle joint relative to the calf; q SW is the posture of the lower leg relative to the world coordinate system W; combined with the above constraints, the equation of the posture quaternion of the foot end relative to the world coordinate system and the Euler angle of the foot end is constructed to obtain the Euler angle of the robot foot end, as follows:

[0030]

[0031]

[0032] Finally, the Euler angle φ of the corrected foot posture in the world coordinate system is obtained i ,θ i , and build a hexapod robot The posture matrix T of the six foot ends of a person dist , as follows:

[0033]

[0034]

[0035] Among them, H i is the plantar posture of the six feet of the hexapod robot, and finally forms a hexapod robot foot-end posture matrix T with a size of 12×18 dist .

[0036] Preferably, step 2 specifically includes the following steps:

[0037] Step 21: Obtain the current center of mass acceleration of the hexapod robot through IMU1 and IMU2 installed on the hexapod robot body and the desired body angular acceleration , construct the state space expression:

[0038]

[0039] Where I3 is the 3×3 identity matrix; p i is the position vector of the hexapod robot’s center of mass relative to the foot end; F i is the ground reaction force acting on the i-th foot, m is the total mass of the hexapod robot; g is the gravitational acceleration vector; I G is the moment of inertia about the center of mass; and They represent the acceleration and translational velocity of the hexapod robot’s center of mass and the angular acceleration of the robot’s body, respectively; A is the center of mass force matrix;

[0040] Step 22: Combine the centroid force matrix A and the hexapod robot foot posture matrix T dist Substituting into the Hessian matrix Λ, the formula is as follows:

[0041] Λ=A T QA+T dist T PT dist (11),

[0042] Among them, Q is the weight matrix of the centroid force matrix A, P is the weight matrix of the foot-end posture matrix, and the Hessian matrix Λ is the updated body state.

[0043] Preferably, step 3 specifically includes the following steps:

[0044] Step 31: Input the Hessian matrix Λ into the quadratic programming model to solve the ground reaction force vector F, as follows:

[0045]

[0046] Where C is the optimized cost function, F is the ground reaction force vector, and its dimension is 18×1;

[0047] Step 32: Introduce the friction tracking constraint model of the six feet into the quadratic programming model, as follows:

[0048]

[0049] Among them, F kiis the matrix of the k-direction in F, where k is x, y, z, and i is the i-th leg of the hexapod robot;

[0050] Step 33: Solve formula (12) using formula (13) as a constraint condition to obtain the optimal ground reaction force distribution F of the six feet;

[0051] Step 34: Convert the ground reaction force at each foot end into the joint force of the corresponding joint through the robot's Jacobian matrix to control the movement of the hexapod robot legs:

[0052]

[0053] Where J is the Jacobian matrix of the hexapod robot, F j is the joint force of the j-th leg of the hexapod robot;

[0054] Step 35: The hexapod robot sends the force of each joint to the corresponding motor, and the motor receives the control instruction and works.

[0055] Preferably, if the ground reaction force distribution F obtained in step 33 cannot satisfy the friction cone constraint, step 3 further includes:

[0056] Step 35: Introduce slack variables into the quadratic programming model i , the modified friction cone constraint is expressed as:

[0057]

[0058] Formula (15) is used as a constraint to solve formula (12) and obtain the optimal ground reaction force distribution F.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The present invention provides a method and system for adaptive terrain walking control of a hexapod robot based on multi-sensor fusion. By combining ground condition perception technology, the robot's adaptability and stability in uneven, rugged or dynamically changing terrain are significantly enhanced. The IMU bias is estimated using a low-pass filter and the acceleration is corrected using the spherical linear interpolation (SLERP) method, which effectively reduces the drift problem in posture estimation. In addition, the ground information obtained by the IMU is fused with the current state information of the robot to form an updated body state, which is used as the input of the virtual model control (VMC), further improving the robustness of the system. The foot end force is optimized through the quadratic programming (QP) algorithm, thereby achieving efficient and stable complex terrain walking capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is a main flow chart according to a preferred embodiment of the present invention;

[0062] Figure 2 Schematic diagram of the foot of the present invention. DETAILED DESCRIPTION

[0063] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0064] A terrain-adaptive walking control method for a hexapod robot based on multi-sensor fusion is provided. An inertial measurement unit (IMU1) and an inertial measurement unit (IMU2) are provided on the sole and handle of each foot of the hexapod robot, respectively. The inertial measurement units are used to measure the acceleration and angular acceleration of the corresponding parts. The method specifically comprises the following steps:

[0065] Step 1: Use IMU1 and IMU2 to collect data, and use low-pass filtering to assist in estimating the bias of IMU1 and IMU2; based on the world coordinate system, calibrate the data collected by IMU1 and IMU2 to obtain the posture information of the sole of the foot;

[0066] Step 2: Integrate the posture information of the sole into the current state of the hexapod robot to form an updated body state;

[0067] Step 3: Input the updated body state into the quadratic programming solver to obtain the optimal ground reaction force distribution matrix F, and adjust the state of each foot based on the body state and the ground reaction force distribution matrix F.

[0068] Preferably, the step 1 specifically includes the following steps:

[0069] Step 11: Use a low-pass filter to estimate the bias of IMU1 and IMU2, as shown below:

[0070]

[0071] where Δω k+1 is the bias of IMU1 and IMU2 at the next moment; Δω k The bias of IMU1 and IMU2 at the current moment, β is the coefficient of the low-pass filter, which is used to balance the weight of the bias of IMU1 and IMU2 at the previous moment and the bias of IMU1 and IMU2 at the current moment; is the angular velocity measurement value at the current moment;

[0072] Step 12: Determine whether the hexapod robot's feet are in a stationary state:

[0073] If the acceleration in the direction of gravity in the world coordinate system measured by the accelerometers in IMU1 and IMU2 is close to the acceleration due to gravity, that is, |ag<0.1, where a is the measured acceleration and g is the acceleration due to gravity, it indicates that the legs of the hexapod robot are currently stationary. The acceleration is corrected as shown in the following formula:

[0074] Δq corr =SLERP(q I ,Δq acc ; α a ) (2),

[0075] Where Δq corr is the corrected quaternion, SLERP() is the spherical linear interpolation function, Δq acc is the quaternion measured by the accelerometers in IMU1 and IMU2, which represents the difference between the gravity direction measured by the accelerometer and the current posture; I is the unit quaternion ([1,0,0,0] T ), indicating a state without any rotation; α a is the interpolation coefficient;

[0076] If the acceleration measured by the accelerometer in IMU1 is not close to the acceleration due to gravity, that is, |ag|>0.1, it indicates that the hexapod robot's foot is in a swinging state. In this case, the current angular velocity is calculated directly using the bias of IMU1 and IMU2 estimated by the low-pass filter, as shown in the following formula:

[0077]

[0078] where ω corr is the corrected angular velocity, is the current angular velocity measured by the gyroscope, Δω k+1 is the bias of IMU1 and IMU2 at the next moment, and then through ω corr Integrate to obtain the Euler angles φ and θ of IMU1 and IMU2 at the current moment, and convert the Euler angles into attitude quaternion Δq corr ;

[0079] Step 13: Calculate the drift of the foot end relative to the world coordinate system used by the robot. The expression is as follows:

[0080]

[0081] where Δq W′W is the quaternion of the world coordinate offset, Δq corr is the corrected drift quaternion, and the two are multiplied to obtain the quaternion Δq of the foot relative to the world coordinate system W ;

[0082] Step 14. Calculate the ankle joint motion as follows:

[0083]

[0084] Among them, q FA is the posture of the foot relative to the ankle joint; q AS is the posture of the ankle joint relative to the calf; q SW is the posture of the lower leg relative to the world coordinate system W;

[0085] Step 15: Using formula (5) as a constraint, construct an equation for the foot-end posture quaternion relative to the world coordinate system and the foot-end Euler angle, and obtain the foot-end Euler angle of the hexapod robot, as follows:

[0086]

[0087]

[0088] Get the Euler angle φ of the corrected foot posture in the world coordinate system i ,θ i , i is the i-th foot;

[0089] Step 16: Construct the posture matrix T of the six legs of the hexapod robot dist , as follows:

[0090]

[0091]

[0092] Among them, H i is the plantar posture of the six feet of the hexapod robot, and finally forms a hexapod robot foot-end posture matrix T with a size of 12×18 dist .

[0093] Step 2 specifically includes the following steps:

[0094] Step 21. The inertial measurement unit (IMU3) installed on the hexapod robot is used to obtain the current state information of the robot. Based on this information, a state space expression is constructed. This expression describes the dynamic characteristics of the robot, including the position, velocity, acceleration, and rotation parameter information of the robot. The state space expression at this time is:

[0095]

[0096] Where I3 is the 3×3 identity matrix; p i is the position vector of the robot foot relative to the robot’s center of mass; F iis the ground reaction force acting on the i-th foot, m is the total mass of the robot; g is the gravitational acceleration vector; I G is the inertia tensor of the robot body; and Represent the acceleration and translational velocity of the robot's center of mass and the angular acceleration of the robot body, respectively. Based on this information, a 6×18 center-of-mass force matrix A is constructed to describe the mass distribution and force distribution of the hexapod robot body.

[0097] Step 22: Combine the centroid force matrix A and the hexapod robot foot posture matrix T dist Substitute into the Hessian matrix Λ, the formula is as follows:

[0098] Λ=A T QA+T dist T PT dist (12),

[0099] Among them, the Hessian matrix Λ is an 18×18 dimensional matrix used to describe the dynamic behavior of the hexapod robot, Q is the weight matrix of the center of mass state matrix, and P is the weight matrix of the foot-end posture matrix. The new state matrix Q and P are obtained based on experience.

[0100] The step 3 specifically includes the following steps:

[0101] The Hessian matrix Λ is input into the quadratic programming (QP) algorithm to solve the ground reaction force vector as follows:

[0102]

[0103] Where C is the optimized cost function, F is the force matrix of the six legs of an 18×1 hexapod robot, and the friction tracking constraint model of the six legs is introduced into the quadratic programming (QP) algorithm as follows:

[0104]

[0105] Among them, the constraint matrix above is a 24×24 constraint matrix, and the F in the matrix ki is the k-direction matrix in F (k is x, y, z, and i is the six legs of the hexapod robot), which is used to describe the physical constraints that must be satisfied by the forces acting on the six legs of the hexapod robot when they come into contact with the ground. Through the above calculations, an optimal ground reaction force distribution F for the six legs is obtained.

[0106] If the reaction force distribution F calculated by quadratic programming (QP) cannot satisfy the friction cone constraint, a slack variable ò is introduced into the constraint matrix in the QP model. i, allowing a certain error range; the modified friction cone constraint can be expressed as:

[0107]

[0108] Adding the constraint condition into formula (13) we can obtain the optimal ground reaction force distribution F;

[0109] The reaction force at each foot end is converted into the joint force of the corresponding joint through the robot's Jacobian matrix to control the movement of the hexapod robot legs:

[0110]

[0111] Where J is the Jacobian matrix of the hexapod robot, F j The hexapod robot sends these joint forces to the corresponding motors. After receiving the instructions, the motors work according to the corresponding instructions.

[0112] At this point, the robot's posture changes due to the forces exerted on its legs. To address this change, the robot uses its IMU sensors to detect the new posture. The sensors collect data and repeat the above steps to adjust the robot's center of gravity. This ensures the robot maintains its balance at all times. The entire control loop operates at a frequency of 100 Hz, allowing the robot to quickly respond to any posture changes and make timely adjustments. This effectively maintains the robot's stability.

[0113] This ground reaction force distribution F achieves the following goals: desired center-of-mass dynamics, ensuring that the robot's actual center-of-mass acceleration and angular acceleration are close to desired values; desired foot-end force direction, ensuring that each leg applies the desired force direction when in contact with the ground; and friction cone constraints, ensuring that the force on each leg is within a reasonable range and complies with the friction cone constraints. This allows the hexapod to control walking on uneven terrain.

[0114] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A terrain adaptive walking control method for a hexapod robot based on multi-sensor fusion, wherein an inertial measurement unit IMU1 and an inertial measurement unit IMU2 are respectively arranged on the sole and the handle of each foot of the hexapod robot, and the inertial measurement unit is used to measure the acceleration and angular acceleration of the corresponding part, characterized in that: The control method specifically comprises the following steps: Step 1: Use IMU1 and IMU2 on each foot to collect data and obtain the foot-end posture matrix of each foot; Step 2: Fuse the foot-end posture matrix into the current state of the hexapod robot to form an updated body state; Step 3: Pass the updated body state information into the quadratic programming solver to obtain the optimal ground reaction force distribution matrix F, and then calculate the joint force F according to the obtained ground reaction force distribution matrix F. j , according to the joint force F j The corresponding joint control instructions are converted and then sent to the joint motors of the hexapod robot to control the movements of the legs of the hexapod robot and keep the body of the hexapod robot stable.

2. The control method according to claim 1, characterized in that: Step 1 specifically includes the following steps: The step 1 specifically includes the following steps: Step 11. Use a low-pass filter to estimate the bias of IMU1 and IMU2 on the same leg, as shown in the following formula: where Δω k+1 is the bias of IMU1 and IMU2 at the next moment; Δω k The bias of IMU1 and IMU2 at the current moment, β is the coefficient of the low-pass filter, which is used to balance the bias of IMU1 and IMU2 at the previous moment and the bias of IMU1 and IMU2 at the current moment; is the angular velocity measurement value at the current moment; Step 12: Determine whether the feet of the hexapod robot are in a stationary state: If the acceleration in the direction of gravity measured by the accelerometers in IMU1 and IMU2 is close to the acceleration of gravity, that is, |ag|<0.1, where a is the measured acceleration and g is the acceleration of gravity, it indicates that the legs of the current hexapod robot are in a stationary state. Spherical linear interpolation (SLERP) is used to correct the acceleration, as shown in the following formula: Δq corr =SLERP(q I ,Δq acc ;α a ) (2), where Δq corr is the corrected quaternion, Δq acc is the quaternion measured by the accelerometers in IMU1 and IMU2, which represents the difference between the gravity direction measured by the accelerometer and the current posture; q I is the unit quaternion ([1,0,0,0] T ), indicating a state without any rotation; α a is the interpolation coefficient, and the current attitude quaternion Δq is obtained corr ; If the accelerometers in IMU1 and IMU2 measure the acceleration in the direction of gravity in the world coordinate system close to the gravity acceleration, that is, |ag|>0.1, it means that the current hexapod robot's feet are in a swinging state. At this time, the current angular velocity is calculated directly using the bias of IMU1 and IMU2 estimated by the low-pass filter, as shown in the following formula: where ω corr is the corrected angular velocity, is the current angular velocity measured by the gyroscope, Δω k+1 is the bias of IMU1 and IMU2 at the next moment, and then through ω corr The current attitude angles φ and θ of IMU1 and IMU2 are obtained by integration, and then the attitude quaternion Δq is obtained by Euler angle quaternion transformation. corr ; Step 13: Calculate the drift of the foot end relative to the world coordinate system used by the robot. The expression is as follows: where Δq W′W is the quaternion of the world coordinate offset, Δq corr is the corrected drift quaternion, and the two are multiplied to obtain the corrected drift quaternion Δq W′ ; Step 14: Calculate the movement of the ankle joint as follows Among them, q FA is the posture of the foot relative to the ankle joint; q AS is the posture of the ankle joint relative to the calf; q SW is the posture of the lower leg relative to the world coordinate system W; combined with the above constraints, the equation of the posture quaternion of the foot end relative to the world coordinate system and the Euler angle of the foot end is constructed to obtain the Euler angle of the robot foot end, as shown in the following formula: Finally, the Euler angle φ of the corrected foot posture in the world coordinate system is obtained i ,θ i and ψ i , and construct the posture matrix T of the six feet of the hexapod robot dist , as follows: Among them, H i is the plantar posture of the six feet of the hexapod robot, and finally forms a hexapod robot foot-end posture matrix T with a size of 12×18 dist .

3. The control method according to claim 1, characterized in that: Step 2 specifically includes the following steps: Step 21: Obtain the current center of mass acceleration of the hexapod robot through IMU1 and IMU2 installed on the hexapod robot body and the expected body angular acceleration , construct the state space expression: Where I3 is the 3×3 identity matrix; p i is the position vector of the hexapod robot’s center of mass relative to the foot end; F i is the ground reaction force acting on the i-th foot, m is the total mass of the hexapod robot; g is the gravitational acceleration vector; I G is the moment of inertia about the center of mass; and They represent the acceleration and translational velocity of the hexapod robot’s center of mass and the angular acceleration of the robot’s body, respectively; A is the center of mass force matrix; Step 22: The centroid force matrix A and the hexapod robot foot posture matrix T dist Substituting into the Hessian matrix Λ, the formula is as follows: Λ=A T QA+T dist T PT dist (11), Among them, Q is the weight matrix of the centroid force matrix A, P is the weight matrix of the foot-end posture matrix, and the Hessian matrix Λ is the updated body state.

4. The control method according to claim 1, characterized in that: The step three specifically includes the following steps: Step 31, input the Hessian matrix Λ into the quadratic programming model to solve the ground reaction force vector F, as follows: Among them, C is the optimized cost function, F is the ground reaction force vector, and the dimension is 18×1; Step 32: Introduce the friction constraint model of six feet into the quadratic programming model, as follows: Among them, F ki is the matrix of the k-direction in F, k is x, y, z, and i is the i-th leg of the hexapod robot; Step 33, using formula (13) as a constraint condition to solve formula (12) to obtain the optimal ground reaction force distribution F of the six feet; Step 34: Convert the ground reaction force at each foot end into the joint force of the corresponding joint through the Jacobian matrix of the robot to control the movement of the six-legged robot leg: F j =J T F (14), Where J is the Jacobian matrix of the hexapod robot, F j is the joint force of the jth leg of the hexapod robot; Step 35: The hexapod robot sends each joint force to the corresponding motor, and the motor receives the control command and works.

5. The control method according to claim 4, characterized in that: If the ground reaction force distribution F obtained in step 33 cannot satisfy the friction cone constraint, step 3 further includes: Step 35: Introduce slack variables into the quadratic programming model i , the modified friction cone constraint is expressed as: Formula (15) is used as a constraint condition to solve formula (12) and obtain the optimal ground reaction force distribution F.

Images