Model predictive control-based humanoid robot obstacle avoidance method
Through laser scanners and depth cameras to detect obstacles, combined with model prediction control and adaptive nonlinear control, humanoid robots achieve stable inverted flips across obstacles, solving the inefficient and instability of existing obstacle avoidance methods in confined spaces The problem of improving mobility efficiency and flexibility.
Patent Information
- Application Number
- CN202411097297.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-08-12
AI Technical Summary
When existing humanoid robots avoid obstacles, side-step obstacle avoidance is inefficient in narrow spaces. Turning and avoiding obstacles increases the length and time of paths, jumping and avoiding obstacles is unstable and posing risks to the robot and the environment.
Adopt obstacle avoidance method based on model prediction control, the road terrain and obstacle height can be detected through laser scanner and depth camera, determine whether it can be reversed and crossed, build a dynamic model and use the model prediction control algorithm for posture adjustment, combined with the self-organized Adapt to nonlinear controllers to ensure stability.
It realizes that directly crossing obstacles without changing the direction of travel, improves movement efficiency, adapts to different obstacle heights and environments, enhances robot flexibility and stability, and reduces the path length and time of traditional obstacle avoidance methods.
Smart Images

Figure CN118915773A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of humanoid robot control, and in particular to a method for humanoid robot obstacle avoidance based on model predictive control. Background Art
[0002] At present, the obstacle avoidance actions adopted by humanoid robots are as follows: 1. Side step obstacle avoidance: that is, the robot avoids obstacles in front by moving its feet to one side; but when it is necessary to quickly pass through a narrow space or avoid an emergency obstacle, side step obstacle avoidance is not as direct and efficient as the method of standing upside down and crossing the obstacle. In extremely narrow environments, side step obstacle avoidance may be limited by space and cannot effectively avoid obstacles; 2. Turn obstacle avoidance: that is, when the robot detects an obstacle in front, it can change the direction of travel by turning around to avoid the obstacle; but turning obstacle avoidance means that the robot needs to change the original direction of travel, which may increase the path length and time to reach the destination, and in some complex environments, turning obstacle avoidance will be limited by the flexibility of the robot or the surrounding obstacles, while the method of standing upside down and crossing the obstacle does not require changing direction, saving time and improving movement efficiency; 3. Jump obstacle avoidance: that is, for some higher obstacles, the robot jumps over them; but this method requires the robot to have a powerful power system and precise jumping control, and has great instability. There may be a risk of losing control or falling during the jumping process, causing damage to the robot and the surrounding environment. Summary of the invention
[0003] The present invention discloses a method for humanoid robot obstacle avoidance based on model predictive control, and the specific method is as follows:
[0004] The laser scanner arranged on the head of the humanoid robot is used to scan the road in front of the humanoid robot, and a coordinate point cloud is formed according to the scanning result to draw a topographic map of the road in front of the humanoid robot;
[0005] The height of the obstacle is calculated by using a depth camera installed on the patella of the humanoid robot;
[0006] Determine whether the obstacle height is greater than the leg length of the humanoid robot;
[0007] If yes, the humanoid robot is controlled to walk to the obstacle and perform an obstacle bypass operation;
[0008] If not, the humanoid robot is controlled to walk to the obstacle and perform an inverted flip over the obstacle.
[0009] Furthermore, a coordinate point cloud is formed, and the specific formula is as follows:
[0010] X=Lcosαcosβ
[0011] Y=Lcosαcosβ
[0012] Z=Lcosαcosβ
[0013] Where L is the distance from the target scanning point to the laser scanner after the emitted laser is reflected back, α and β are the rotation angles of the laser reflector of the laser scanner in the horizontal and vertical directions, respectively, and X, Y, and Z are the x, y, and z axis coordinates of the target scanning point relative to the laser scanner, respectively.
[0014] Furthermore, the height of the obstacle is calculated using the following formula:
[0015]
[0016] Where Z is the distance between the imaging point at the bottom of the obstacle and the depth camera mounted on the hip of the humanoid robot, f is the focal length of the depth camera, T is the distance between the imaging point at the top of the obstacle and the depth camera mounted on the hip of the humanoid robot, and x is 0l x is the horizontal coordinate of the scanning point of the obstacle in front of the image on the left camera of the depth camera, 0r It is the horizontal coordinate of the scanning point of the obstacle in front of the image on the right camera of the depth camera.
[0017] Furthermore, the inverted flip over obstacle operation is performed, and the specific method is as follows:
[0018] Construct a dynamic model of a humanoid robot;
[0019] A humanoid robot controller based on the model predictive control algorithm is used to complete the posture adjustment of the humanoid robot during the inverted flip stride action.
[0020] Furthermore, a humanoid robot dynamics model is constructed. The specific method is as follows:
[0021] The humanoid robot is simplified as a single-mass inverted pendulum with a spring and a damper, and the equation of motion is:
[0022]
[0023] Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, τ is the torque applied to the robot, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction to the ground, and u is δ is the output angle of the inverted pendulum, which is input by the robot controller;
[0024] u is the displacement of the controller relative to the center of mass, which is approximately:
[0025]
[0026] Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, τ is the torque applied to the robot, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction to the ground, and u is δ is the output angle of the inverted pendulum, which is input by the robot controller;
[0027] The zero moment point ZMP of the robot in the x-direction is calculated as follows:
[0028]
[0029] In the formula, F y is the ground reaction force in the y direction, F z is the ground reaction force in the z direction;
[0030] Because F y It is also the reaction torque of the spring and shock absorber, which can be written as:
[0031]
[0032] Where s is the length from the support point to the mass point of the inverted pendulum, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, and δ is the angle of the inverted pendulum, which corresponds to the inclination angle of the robot's legs in the vertical direction of the ground;
[0033] When undisturbed, it can be deduced that and The state space equations of the two formulas are:
[0034]
[0035] Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction of the ground, and u is the displacement of the controller to the center of mass; the two states are the angle δ and angular velocity of the robot
[0036] Further, the humanoid robot controller encodes the distance input signal measured by the depth camera into a neuron, and after the input distance is presented, the neuron that best matches the input distance is selected as the optimal neuron;
[0037] The weights of the optimal neurons are then updated, and under the guidance of the reward function, the correct control action is autonomously found using the input pattern information and the performance indicators that need to be optimized.
[0038] Furthermore, the specific setting method of the humanoid robot controller is as follows:
[0039] Define the network structure and hyperparameters, and randomly initialize the input and output weights;
[0040] Present the input pattern at time I(t) and use the depth camera to measure the distance and select the optimal neuron N S , which connects I(t) with N S The distance between the input weight vectors to the input nodes is minimized; where I(t) is the input pattern;
[0041] For the optimal neuron N S , the corresponding output weight Used to generate control actions; that is, the orbiting version At the optimal neuron weights Add a Gaussian zero-mean random variable λ to the optimal neuron N S The associated gain α S ; α S Starting from an initial value and fixing a final value, α S It decreases linearly between these two values; where O(t) is the output control signal;
[0042] The actual reward function R(t) is evaluated together with the reward increment ΔR=R(t)-R(t-1); if ΔR>b s , where b s For the neuron N S The average increment of the correlation; b s The update rules are as follows:
[0043]
[0044] Where σ>0, is the new average increment, is the old average increment, ΔR is the average increment;
[0045] Connect to the optimal neuron N S The input and output weights of and its neighboring neurons are updated according to the following rules:
[0046] w i,in (t+1)=w i,in (t)+η in ζ(I(t)-w i,in (t))
[0047] w i,out (t+1)=w i,out (t)+η out ζ(O(t)-w i,out (t))
[0048] where η in is the input learning rate, η out is the output learning rate, w i,in Input weight, w i,out is the output weight, ζ is the neighborhood function of the neuron, if N R Represents a neighborhood with a radius of R, then in the current optimal neuron N S Around, there are:
[0049]
[0050] ζ established a S The radius neighborhood, the weight is updated within this range;
[0051] In each learning iteration, given an output pattern, an optimal neuron N is selected S And provide an output signal, then evaluate the reward and perform the learning phase, that is, update w i,in , w i,out , b s and α S , after several iterations, updates of other learning hyperparameters are performed.
[0052] Furthermore, after the inverted flip over obstacle maneuver, the humanoid controller calculates the forces required to maintain the body standing position, and then uses these forces as input signals to the controller, which provides stable joint torques;
[0053] If the robot structure changes and causes the center of gravity to shift, an adaptive nonlinear controller MMC is added as a feedforward error compensator to adjust the robot's posture. This adaptive nonlinear controller MMC provides an additional nonlinear feedforward control action, which is added to the underlying leg joint controller in the form of force after transforming the joint torque through the Jacobian matrix; a reward function is introduced:
[0054]
[0055] Where Θ ref is the reference pitch velocity of the robot, Θ is the actual pitch velocity of the robot, is the reference pitch acceleration of the robot, is the actual pitch acceleration of the robot, β is the introduced gain, and the priority of pitch velocity to pitch position error is established: as the pitch error decreases, the relevance of the pitch velocity error becomes greater;
[0056] The algorithm that controls the robot updates the output weights to ensure the stability of the stride by improving the reward function, which takes into account the error between the reference and the actual pitch of the robot.
[0057] Due to the adoption of the above technical solution, the present invention has the following beneficial effects:
[0058] 1. The forward inverted method of crossing obstacles can directly cross obstacles without having to go around or avoid them like other methods. This avoids the situation where traditional obstacle avoidance methods increase the moving distance and time, and can significantly improve movement efficiency in specific scenarios.
[0059] 2. The method of crossing obstacles by front handstand can adapt to different types of obstacles below the height of the legs by adjusting the height and posture of the handstand, and has a certain degree of flexibility. When traditional obstacle avoidance methods are difficult to effectively deal with in rugged or narrow spaces, this method can overcome these limitations and achieve stable movement.
[0060] 3. The model prediction algorithm introduced in the present invention can predict the motion trajectory in the future according to the current system state and the prediction model, and has strong robustness and flexibility; and an adaptive nonlinear controller is introduced that does not require a priori knowledge of the robot's physical characteristics, and the controller does not require internal iteration. After properly mapping the input signal, it can quickly reach the maximum value of the reward function, providing endless learning opportunities;
[0061] 4. The present invention combines the method of standing forward and crossing obstacles with a model-based prediction algorithm, which improves work efficiency and stability while completing specific environmental tasks and has broad application prospects.
[0062] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0063] The accompanying drawings of the present invention are as follows.
[0064] Figure 1 It is a control flow diagram of the present invention.
[0065] Figure 2 Schematic diagram of the control loop for the MPC algorithm learning iteration. DETAILED DESCRIPTION
[0066] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0067] A humanoid robot obstacle avoidance method based on model predictive control, such as Figure 1 As shown, the specific steps are as follows:
[0068] S1. Scan the road ahead of the humanoid robot by means of a laser scanner arranged on the head of the humanoid robot, form a coordinate point cloud according to the scanning result, and draw a topographic map of the road ahead of the humanoid robot.
[0069] In step S1, a laser scanner installed on the head of the humanoid robot is used to scan the road surface in front of the humanoid robot and receive the reflected light signal. The generator can calculate the distance between the laser scanner on the humanoid robot and the front scanning point based on the collected reflection data of the light signal, and form a coordinate point cloud map based on the coordinates of multiple front scanning points. The calculation formula of the coordinate point cloud is as follows:
[0070] X=Lcosαcosβ
[0071] Y=Lcosαcosβ
[0072] Z=Lcosαcosβ
[0073] Where L is the distance from the target scanning point to the laser scanner after the emitted laser is reflected back, α and β are the rotation angles of the laser reflector of the laser scanner in the horizontal and vertical directions, respectively, and X, Y, and Z are the x, y, and z axis coordinates of the target scanning point relative to the laser scanner, respectively.
[0074] S2. Calculate the height of the obstacle by using a depth camera installed on the patella of the humanoid robot.
[0075] In step S2, the humanoid robot walks to the obstacle and stops, and the height of the obstacle is calculated by the depth camera mounted on its hip. Specifically, the depth camera uses its left and right cameras to image, and sends the data measured by the depth camera to the main controller. The main controller uses the triangulation principle to calculate and obtain the distance between the imaging point at the bottom of the obstacle and the imaging point at the top of the obstacle in the depth camera as the height of the obstacle. The calculation formula is as follows:
[0076]
[0077] Where Z is the distance between the imaging point at the bottom of the obstacle and the depth camera mounted on the hip of the humanoid robot, f is the focal length of the depth camera, in pixels, T is the distance between the imaging point at the top of the obstacle and the depth camera mounted on the hip of the humanoid robot, and x is 0l x is the horizontal coordinate of the scanning point of the obstacle in front of the image on the left camera of the depth camera, 0rThe horizontal coordinate of the scanning point of the obstacle in front of us is imaged on the right camera of the depth camera. After measuring the height of the obstacle, it is compared with the leg length of the humanoid robot to carry out the subsequent steps.
[0078] The height of the obstacle measured by the depth camera and the coordinate point cloud formed by the laser scanner can be used to obtain the terrain map of the road ahead, which facilitates the robot's next choice of bypassing or standing upside down to cross the obstacle.
[0079] S3. Determine whether the height of the obstacle is greater than the leg length of the humanoid robot.
[0080] If so, the humanoid robot is controlled to walk to the obstacle and perform an operation of bypassing the obstacle.
[0081] If not, the humanoid robot is controlled to walk to the obstacle and perform an inverted flip over the obstacle.
[0082] In step S3, the obstacle height measured by the depth camera using the triangulation principle is compared with the leg length of the humanoid robot; if the obstacle height is greater than the leg length of the humanoid robot, the humanoid robot cannot cross the obstacle by standing upside down, and a detour operation is taken, that is, the main controller plans the footstep sequence of the humanoid robot to bypass the obstacle and return to the original path according to the front terrain map drawn by the laser scanner; if the obstacle height is less than the leg length of the humanoid robot, the humanoid robot can cross the obstacle by standing upside down to complete the obstacle avoidance; the humanoid robot is controlled by deriving a dynamic model: the humanoid robot is modeled and simplified as a single-mass inverted pendulum with a spring and a damper, and the motion equation is:
[0083]
[0084] Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, τ is the torque applied to the robot, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction to the ground, and u is δ is the output angle of the inverted pendulum, which is input by the robot controller;
[0085] Therefore, u is the displacement of the controller relative to the center of mass, which is approximately:
[0086]
[0087] Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, τ is the torque applied to the robot, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction to the ground, and u is δ is the output angle of the inverted pendulum, which is input by the robot controller;
[0088] The zero moment point (ZMP) of the robot in the x-direction is calculated as follows:
[0089]
[0090] In the formula, F y is the ground reaction force in the y direction, F z is the ground reaction force in the z direction;
[0091] Because F y It is also the reaction torque of the spring and shock absorber, which can be written as:
[0092]
[0093] Where s is the length from the support point to the mass point of the inverted pendulum, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, and δ is the angle of the inverted pendulum, which corresponds to the inclination angle of the robot's legs in the vertical direction of the ground;
[0094] When the system model is not disturbed, it can be deduced that
[0095] and The state space equations of the two formulas are:
[0096]
[0097] Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction of the ground, and u is the displacement of the controller to the center of mass; the two states are the angle δ and angular velocity of the robot
[0098] After completing the dynamic modeling of the robot and determining that the robot can cross the obstacle by standing upside down, the model predictive control MPC algorithm is introduced to complete the posture adjustment of the robot in the crossing action, that is, the fast calculation based on the dynamic model of the robot is used to predict the system output within a limited time range, and optimize according to the control action to minimize the specified performance indicators; the MPC algorithm encodes the distance input signal measured by the depth camera into a neuron, which is achieved by considering a learning rule. After the input distance is presented, the neuron that best matches the input distance is selected as the optimal neuron; then the weight of the optimal neuron is updated, and the correct control action is autonomously found under the guidance of the reward function using the input pattern information and the performance indicators to be optimized;
[0099] Below are the steps for designing a humanoid robot controller and applying the algorithm:
[0100] S31. Define the network structure and hyperparameters, and randomly initialize the input and output weights;
[0101] S32, present the input pattern at time I(t), measure the distance using the depth camera, and select the optimal neuron N S , which connects I(t) with N S The distance between the input weight vectors to the input nodes is minimized; where I(t) is the input pattern;
[0102] S33. For the optimal neuron N S , the corresponding output weight Used to generate control actions; that is, the orbiting version At the optimal neuron weights Add a Gaussian zero-mean random variable λ to the optimal neuron N S The associated gain α S ; α S Starting from an initial value and fixing a final value, α S It decreases linearly between these two values; where O(t) is the output control signal;
[0103] S34. The actual reward function R(t) is evaluated together with the reward increment ΔR=R(t)-R(t-1); if ΔR>b s , where b s For the neuron N S The average increment of the correlation; b s The update rules are as follows:
[0104]
[0105] Where σ>0, is the new average increment, is the old average increment, ΔR is the average increment;
[0106] Connect to the optimal neuron N S The input and output weights of and its neighboring neurons are updated according to the following rules:
[0107] w i,in (t+1)=w i,in (t)+η in ζ(I(t)-w i,in (t))
[0108] w i,out (t+1)=w i,out (t)+η out ζ(O(t)-w i,out (t))
[0109] where η in is the input learning rate, η out is the output learning rate, w i,in Input weight, w i,out is the output weight, ζ is the neighborhood function of the neuron, if N R Represents a neighborhood with a radius of R, then in the current optimal neuron N S Around, there are:
[0110]
[0111] So ζ builds a S The radius neighborhood, the weight is updated within this range;
[0112] In each learning iteration, given an output pattern, an optimal neuron N is selected S And provide an output signal, then evaluate the reward and perform the learning phase, that is, update w i,in , w i,out , b s and α S , after several iterations, perform updates of other learning hyperparameters; Figure 2 is the control loop flow chart for the i-th learning iteration;
[0113] S4. After completing the leaping action, the MPC algorithm calculates the forces required to maintain the standing position of the body, and then uses these forces as input signals for the controller, which provides stable joint torques. If the robot structure changes and causes the center of gravity to shift, an adaptive nonlinear controller MMC is added as a feedforward error compensator to adjust the robot's posture. This adaptive nonlinear controller MMC provides an additional nonlinear feedforward control action, which is added to the underlying leg joint controller in the form of force after transforming the joint torque through the Jacobian matrix. A reward function is introduced:
[0114]
[0115] Where Θ ref is the reference pitch velocity of the robot, Θ is the actual pitch velocity of the robot, is the reference pitch acceleration of the robot, is the actual pitch acceleration of the robot, β is the introduced gain, and the priority of pitch velocity to pitch position error is established: as the pitch error decreases, the relevance of the pitch velocity error becomes greater;
[0116] The algorithm for controlling the robot updates the output weights by improving the reward function to ensure the stability of the striding action. For simplicity, only the pitch angle of the sagittal plane is limited. The control goal is to keep the pitch angle of the robot as close to the reference angle as possible, so the reward function takes into account the error between the reference and the actual pitch of the robot.
[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A humanoid robot obstacle avoidance method based on model predictive control, It is characterized in that The specific method is as follows: The laser scanner arranged on the head of the humanoid robot is used to scan the road in front of the humanoid robot, and a coordinate point cloud is formed according to the scanning result to draw a topographic map of the road in front of the humanoid robot; The height of the obstacle is calculated by using a depth camera installed on the patella of the humanoid robot; Determine whether the obstacle height is greater than the leg length of the humanoid robot; If yes, the humanoid robot is controlled to walk to the obstacle and perform an obstacle bypassing operation; If not, the humanoid robot is controlled to walk to the obstacle and perform an inverted flip over the obstacle.
2. The method for avoiding obstacles of a humanoid robot based on model predictive control as claimed in claim 1, It is characterized in that The coordinate point cloud is formed. The specific formula is as follows: X=Lcosαcosβ Z=Lcosαcosβ Where L is the distance from the target scanning point to the laser scanner after the emitted laser is reflected back, α and β are the rotation angles of the laser reflector of the laser scanner in the horizontal and vertical directions, respectively, and X, Y, and Z are the x, y, and z axis coordinates of the target scanning point relative to the laser scanner, respectively.
3. The method for avoiding obstacles of a humanoid robot based on model predictive control as claimed in claim 1, It is characterized in that Calculate the height of the obstacle using the following formula: Where Z is the distance between the imaging point at the bottom of the obstacle and the depth camera mounted on the hip of the humanoid robot, f is the focal length of the depth camera, T is the distance between the imaging point at the top of the obstacle and the depth camera mounted on the hip of the humanoid robot, and x is 0l is the horizontal coordinate of the scanning point of the obstacle in front of us imaged on the left camera of the depth camera, x 0r It is the horizontal coordinate of the scanning point of the obstacle in front of the image on the right camera of the depth camera.
4. The method for avoiding obstacles of a humanoid robot based on model predictive control as claimed in claim 1, It is characterized in that To perform an inverted flip over an obstacle, the specific steps are as follows: Construct a dynamic model of a humanoid robot; A humanoid robot controller based on the model predictive control algorithm is used to complete the posture adjustment of the humanoid robot during the inverted flip stride action.
5. The method for avoiding obstacles of a humanoid robot based on model predictive control as claimed in claim 4, It is characterized in that Construct a humanoid robot dynamics model. The specific method is as follows: The humanoid robot is simplified as a single-mass inverted pendulum with a spring and a damper, and the equation of motion is: Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, τ is the torque applied to the robot, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction to the ground, and u is δ is the output angle of the inverted pendulum, which is input by the robot controller; u is the displacement of the controller relative to the center of mass, which is approximately: Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, τ is the torque applied to the robot, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction to the ground, and u is δ is the output angle of the inverted pendulum, which is input by the robot controller; The zero moment point ZMP of the robot in the x-direction is calculated as follows: In the formula, F y is the ground reaction force in the y direction, F z is the ground reaction force in the z direction; Because F y It is also the reaction torque of the spring and shock absorber, which can be written as: Where s is the length from the support point to the mass point of the inverted pendulum, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, and δ is the angle of the inverted pendulum, which corresponds to the inclination angle of the robot's legs in the vertical direction of the ground; When undisturbed, it can be deduced that and The state space equations of the two formulas are: Where m is the mass of the robot, s is the length from the support point of the inverted pendulum to the mass point, that is, the height from the ground to the mass point when the robot is in the preparation posture before walking, c is the damping constant, k is the spring constant, g is the acceleration of gravity, δ is the angle of the inverted pendulum, corresponding to the inclination angle of the robot's legs in the vertical direction of the ground, and u is the displacement of the controller to the center of mass; the two states are the angle δ and angular velocity of the robot 6. The method for avoiding obstacles of a humanoid robot based on model predictive control as claimed in claim 5, It is characterized in that The humanoid robot controller encodes the distance input signal measured by the depth camera into a neuron, and after the input distance is presented, the neuron that best matches the input distance is selected as the optimal neuron; The weights of the optimal neurons are then updated, and under the guidance of the reward function, the correct control action is autonomously found using the input pattern information and the performance indicators that need to be optimized.
7. The method for avoiding obstacles of a humanoid robot based on model predictive control as claimed in claim 6, It is characterized in that Humanoid robot controller, the specific setting method is as follows: Define the network structure and hyperparameters, and randomly initialize the input and output weights; Present the input pattern at time I(t) and use the depth camera to measure the distance and select the optimal neuron N S , which connects I(t) with N S The distance between the input weight vectors to the input nodes is minimized; where I(t) is the input pattern; For the optimal neuron N S , the corresponding output weight Used to generate control actions; that is, the orbiting version At the optimal neuron weights Add a Gaussian zero-mean random variable λ to the optimal neuron N S The associated gain α S ; α S Starting from an initial value and fixing a final value, α S It decreases linearly between these two values; where O(t) is the output control signal; The actual reward function R(t) is evaluated together with the reward increment ΔR=R(t)-R(t-1); if ΔR>b s , where b s For the neuron N S The average increment of the correlation; b s The update rules are as follows: Where σ>0, is the new average increment, is the old average increment, ΔR is the average increment; Connect to the optimal neuron N S The input and output weights of and its neighboring neurons are updated according to the following rules: where η in is the input learning rate, η out is the output learning rate, w i,in Input weight, w i,out is the output weight, is the neighborhood function of the neuron, if N R Represents a neighborhood with a radius of R, then in the current optimal neuron N S Around, there are: Built around N S The radius neighborhood, the weight is updated within this range; In each learning iteration, given an output pattern, an optimal neuron N is selected S And provide an output signal, then evaluate the reward and perform the learning phase, that is, update w i,in , w i,out , b s and α S , after several iterations, updates of other learning hyperparameters are performed.
8. The method for avoiding obstacles of a humanoid robot based on model predictive control as claimed in claim 7, It is characterized in that After the inverted flip over obstacle maneuver, the humanoid controller calculates the forces required to maintain the body in a standing position, and then uses these forces as input signals to the controller, which provides stable joint torques; If the robot structure changes and causes the center of gravity to shift, an adaptive nonlinear controller MMC is added as a feedforward error compensator to adjust the robot's posture. This adaptive nonlinear controller MMC provides an additional nonlinear feedforward control action, which is added to the underlying leg joint controller in the form of force after transforming the joint torque through the Jacobian matrix; a reward function is introduced: Where Θ ref is the reference pitch velocity of the robot, Θ is the actual pitch velocity of the robot, is the reference pitch acceleration of the robot, is the actual pitch acceleration of the robot, β is the introduced gain, and the priority of pitch velocity to pitch position error is established: as the pitch error decreases, the relevance of the pitch velocity error becomes greater; The algorithm that controls the robot updates the output weights to ensure the stability of the stride by improving the reward function, which takes into account the error between the reference and the actual pitch of the robot.
Citation Information
Patent Citations
Pre-planning type real-time gait control algorithm
CN109333534A
Control system for gait walking planning of small biped robot
CN109870947A
Stable walking control method and device for biped robot, equipment and readable medium
CN111674486A
Autonomous navigation method and device for quadruped robot, computer equipment and storage medium
CN111752285A
Humanoid robot modeling method and device based on floating base flywheel inverted pendulum
CN113119102A
Cited By
Humanoid robot complex terrain balance control method based on state and disturbance observer
CN119575810A
Multi-degree-of-freedom robot control method and system
CN120347783A
A multi-degree-of-freedom robot control method and system
CN120347783B
Whole body obstacle avoidance method and system for a legged humanoid robot
CN123010811A
