A gait switching control method and system for a legged robot

Abstract

The present application belongs to the technical field of robot control, and more particularly relates to a gait switching control method and system for a legged manipulator robot, which comprises the following steps: calculating the current gait of the legged manipulator robot and the positions of the zero moment points of the gait to be switched; planning a zero moment point reference trajectory for the legged manipulator robot; using a model predictive control algorithm to control the legged manipulator robot to move according to the zero moment point reference trajectory, and adjusting the control parameters of the predictive control algorithm in real time according to a loss function, while outputting a state trajectory and an input trajectory in real time, so as to realize gait switching of the legged manipulator robot. The gait switching control method for the legged manipulator robot provided by the present application fully considers the influence of manipulator dynamics on the motion performance of the whole robot, and can realize gait switching control of the legged robot.

Description

Technical Field

[0001] This invention belongs to the technical field of robot control technology, and more specifically, relates to a gait switching control method and system for a legged robotic arm robot. Background Technology

[0002] Legged robots can move by using discrete footholds, giving them a significant advantage in environmental adaptability compared to wheeled robots. The robotic arms mounted on quadruped robots, in addition to manipulating objects, can also assist in balancing the robot, improving its dynamic movement performance. In certain specific situations, such as space-constrained environments or when leg joints are damaged, quadruped robots need to rely on two legs to maintain balance and mobility.

[0003] However, coordinating the movement of the quadruped robot and the robotic arm to enable the robot to move on two legs is a challenging problem. Specifically, it involves controlling the legged robotic arm robot to enter a bipedal standing state and maintain balance in the bipedal standing state. Existing technologies lack gait switching control methods for quadruped robotic arms. Summary of the Invention

[0004] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides a gait switching control method and system for a legged robotic arm robot, realizing the switching between quadrupedal gait and bipedal gait of the quadruped robotic arm robot.

[0005] To achieve the above objectives, according to one aspect of the present invention, a gait switching control method for a legged robotic arm is provided, the method comprising the following steps:

[0006] S1 calculates the position of the zero-moment point of the current gait and the gait to be switched of the legged robotic arm robot, respectively;

[0007] S2 plans the motion trajectory of the zero-moment point of the legged robotic arm robot from the current gait and the gait to be switched to the gait based on the position and linear velocity of the zero-moment point of the current gait and the time from the current gait to the gait to be switched. This motion trajectory serves as the reference trajectory of the zero-moment point.

[0008] S3 uses a predictive control algorithm to control the legged robotic arm robot to move according to the zero-moment point reference trajectory. It calculates the tracking error between the actual motion trajectory and the zero-moment point reference trajectory in real time, uses the tracking error to calculate the loss function of the predictive control algorithm, and adjusts the control parameters of the predictive control algorithm in real time according to the loss function. At the same time, it outputs the state trajectory and the input trajectory in real time, thereby realizing the gait switching of the legged robotic arm robot. The predictive control algorithm takes the zero-moment point reference trajectory as input and the state trajectory and the input trajectory as output.

[0009] More preferably, the loss function expression obtained in step S3 is:

[0010]

[0011] Where x and u represent the state trajectory and input trajectory, respectively, and P(x,t) is the position of the robot's zero-torque point at the current time t. ref For the zero-moment point reference trajectory, the positive definite matrix Q and the positive semi-definite matrix R are the weight matrices for tracking the reference state trajectory and the input trajectory, respectively. P It tracks the reference trajectory P(t) at the zero torque point. ref The weight matrix.

[0012] More preferably, in step S3, the predictive control algorithm needs to satisfy the following constraints, floating basis dynamics constraints:

[0013]

[0014] in, The linear and angular momentum at the center of mass of the legged robotic arm represent the linear and angular momentum. Representing generalized coordinates, including the six-dimensional coordinates q of the robot's base coordinate system in the inertial frame. base q leg q represents the leg joint angles of a quadruped robot. arm A represents the joint angle of the robotic arm. b and A j Represents the center-of-mass momentum matrix A com The corresponding parts of the fuselage and the corresponding parts of the joints, f ci and These represent the vector pointing from the center of mass to the contact point and the three-dimensional contact force exerted by the ground on the robot, respectively.

[0015] When a legged robot makes contact, the foot tip must satisfy a no-slip constraint with respect to the ground.

[0016]

[0017] In the non-contact phase, the trajectory of the foot's swing point must follow the normal vector of the ground. And track the reference trajectory v * (t), when in the non-contact term, the contact force is zero:

[0018]

[0019] f ci =0

[0020] When the foot touches the ground, the contact force must satisfy the friction cone constraint:

[0021]

[0022] Where μ is the ground friction coefficient. and These represent the components of the ground contact force in the three directions.

[0023] More preferably, in step S3, the state trajectory includes the linear momentum and angular momentum at the center of mass of the legged robotic arm in the current state, the joint angles of the legs, and the joint angles of the robotic arm; the input trajectory includes the three-dimensional contact force of the ground on the legged robotic arm, the joint velocities of the legs, and the joint velocities of the robotic arm.

[0024] More preferably, by redistributing the robot's contact forces and joint torques through hierarchical quadratic programming, the robot tracks an optimized trajectory to meet dynamic constraints and target tracking tasks.

[0025] The dynamic constraint tasks include floating base dynamic constraints, joint moment limitation constraints, contact foot no-slip constraints, and friction cone constraints:

[0026]

[0027] Target tracking tasks include swing foot tracking, floating base, robotic arm joint commands, and contact foot force tracking tasks:

[0028]

[0029] J ci This represents the contact point c. i The floating base of the bicorn at that location, and These represent the stiffness and damping coefficients, respectively, for tracking the foot's swing trajectory. and These represent the optimized foot joint angles. Calculate the desired position and velocity of the foot. and These represent the actual position and speed of the foot, respectively. and The stiffness and damping coefficients represent the end effector's trajectory tracking. and This represents the current joint angle and joint speed of the robotic arm.

[0030] More preferably, the specific calculation method for the zero torque point position in step S1 is as follows:

[0031]

[0032] Where x cog y cog and zcog p represents the coordinates of the robot's center of gravity in various directions in the inertial frame. x p y and p z It is a symbolic representation of the three-dimensional coordinates of the zero-torque point.

[0033] More preferably, a Bézier curve is used to fit and generate the zero-moment point reference trajectory. The formula for generating the zero-moment point reference trajectory is as follows:

[0034] P(τ)=(1-τ) 3 A+3(1-τ) 2 τC+3(1-τ)τ 2 D+τ 3 B

[0035]

[0036] t=τ·T traj

[0037] In this context, control points A and B represent the initial and final two-dimensional coordinates of the zero-torque point, while control points C and D are located along the initial velocity direction of the zero-torque point, respectively. and final velocity direction and Representing the initial and final velocity values, the parameter τ∈[0,1] represents the parameters of the entire process from the initial point P1 to the final point P2; adding a time scale to the Bézier curve: t=τ·T traj Among them, T traj This represents the total duration of the zero-torque point trajectory from the initial point P1 to the final point P2.

[0038] According to a second aspect of the present invention, a gait switching control system for a legged robotic arm that implements any of the above-described gait switching control methods is provided, the gait switching control system comprising:

[0039] The calculation module is used to calculate the zero-torque point position of the legged robotic arm robot;

[0040] Zero-moment point planning module, which is used to plan the zero-moment point reference trajectory of the legged robotic arm robot from the current gait to the gait to be switched;

[0041] The predictive control module is used to construct a predictive control algorithm with a zero-torque point reference trajectory as input and a state trajectory and an input trajectory as output. The predictive control algorithm is used to control the movement of the legged robotic arm robot and outputs the state trajectory and input trajectory in real time to control the gait switching of the legged robotic arm robot.

[0042] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art:

[0043] This invention considers the influence of robotic arm dynamics on the overall motion performance of the robot, and proposes a gait switching control method and system for a legged robotic arm robot based on zero-torque point trajectory planning. This method assists the legged robot in completing bipedal gait switching and enables stable bipedal gait switching and walking of a quadruped robot.

[0044] This invention employs a nonlinear model predictive control algorithm based on center-of-mass dynamics to track the zero-moment reference trajectory and adds the tracking error value to the loss function of the model predictive control. By combining appropriate constraints, including floating basis dynamics constraints, no-slip constraints, contact force constraints, and friction cone constraints, the state trajectory and input trajectory are optimized to ensure successful gait switching and maintain the balance and stability of the legged robotic arm. The control framework proposed in this invention can be extended to various gait movements to cope with extreme scenarios such as robot joint damage. Attached Figure Description

[0045] Figure 1 This is a structural schematic diagram of a quadrupedal robotic arm robot constructed according to a preferred embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of a control framework constructed according to a preferred embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of the trajectory from quadrupedal gait to bipedal gait constructed according to a preferred embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0049] This application provides a gait switching control method for a quadrupedal robotic arm robot. The method includes the following steps: S1 Calculate the position of the zero-moment point of the current gait and the gait to be switched of the quadrupedal robotic arm robot; S2 Based on the position and linear velocity of the zero-moment point of the current gait and the gait to be switched, as well as the time from the current gait to the gait to be switched, plan the motion trajectory of the zero-moment point of the quadrupedal robotic arm robot from the current gait to the gait to be switched, and use this motion trajectory as the zero-moment point reference trajectory; S3 Construct a predictive control algorithm with the zero-moment point reference trajectory as input and the state trajectory and input trajectory as output. Use the predictive control algorithm to control the quadrupedal robotic arm robot to move according to the zero-moment point reference trajectory, and calculate the tracking error between the actual motion trajectory and the zero-moment point reference trajectory in real time. Use the tracking error to calculate the loss function of the predictive control algorithm, adjust the control parameters of the predictive control algorithm in real time according to the loss function, and output the state trajectory and input trajectory in real time, thereby realizing the gait switching of the quadrupedal robotic arm robot.

[0050] In one possible embodiment, see [reference] Figure 1 and Figure 2 As shown, the quadrupedal robotic arm consists of a 12-DOF quadruped robot and a 6-DOF robotic arm. The quadruped robot has four legs, each with three joints, driven by motors. The joints of each leg are the hip joint, knee joint, and ankle joint. The robotic arm is mounted on the robot's body and contains six DDOF joints, used to perform various maneuvers such as grasping, carrying, and operating switches.

[0051] In one possible embodiment, the trajectory of the zero-moment point (ZMP) is planned based on the different ranges of motion for quadrupedal and bipedal gaits, see [reference]. Figure 3 As shown, in quadrupedal gait, ZMP moves within the support polygon formed by the four legs, while in bipedal gait, ZMP moves within the support line segment formed by the two legs.

[0052] To achieve a stable transition from quadrupedal gait to bipedal gait, it is necessary to ensure a smooth transition of the ZMP trajectory during the transition. The specific steps are as follows: Calculate the ZMP position of the current quadrupedal gait:

[0053]

[0054] Where x cog y cog and z cog p represents the coordinates of the robot's center of gravity in various directions in the inertial frame. x p y and p z This is the symbolic representation of ZMP three-dimensional coordinates. Specifically, when the ground is a plane, p... z=0

[0055] Plan the ZMP trajectory from the center of the support polygon in quadrupedal gait to the center of the support line segment in bipedal gait. Fit the ZMP trajectory using a Bézier curve to ensure a smooth transition. See [link to ZMP transition process] for details. Figure 3 The ZMP reference trajectory is generated using the following formula:

[0056] P(τ)=(1-τ) 3 A+3(1-τ) 2 τC+3(1-τ)τ 2 D+τ 3 B

[0057]

[0058] In this context, control points A and B represent the initial and final two-dimensional coordinates of the ZMP, while control points C and D are located along the initial velocity direction of the ZMP. and final velocity direction and P(t) represents the initial and final velocity values, P(t) represents the ZMP reference trajectory, and the parameter τ∈[0,1] represents the parameters of the entire process from the initial point P1 to the final point P2.

[0059] Add a time scale to the above Bézier curve, t = τ·T traj

[0060] Among them, T traj This represents the total duration of the ZMP reference trajectory from the initial point P1 to the final point P2.

[0061] To track the planned ZMP reference trajectory P(t), this embodiment uses a nonlinear model predictive control algorithm to track the reference state trajectory x. ref Input trajectory u ref And the ZMP reference trajectory P(t) ref The operating loss function of model predictive control is shown in the following equation:

[0062]

[0063] Where P(x,t) is the robot's position at the current time t in ZMP, and the positive definite matrix Q and the semi-positive definite matrix R are the weight matrices for tracking the reference state trajectory and the input trajectory, respectively. P It tracks the ZMP reference trajectory P(t). ref The weight matrix.

[0064] To ensure that the control strategy of the legged robotic arm does not violate the laws of physics, model predictive control must meet certain constraints. First, it must satisfy the floating basis dynamics constraints.

[0065]

[0066] in, The linear and angular momentum at the center of mass of the legged robotic arm represent the linear and angular momentum. Representing generalized coordinates, including the six-dimensional coordinates q of the robot's base coordinate system in the inertial frame. base q leg q represents the leg joint angles of a quadruped robot. arm This represents the joint angles of the robotic arm. A b and A j Represents the center-of-mass momentum matrix A com The corresponding parts of the fuselage and the corresponding parts of the joints. ci and These represent the vector pointing from the center of mass to the contact point and the three-dimensional contact force exerted by the ground on the robot, respectively.

[0067] When a legged robot makes contact, the foot tip must satisfy a no-slip constraint with respect to the ground.

[0068]

[0069] In the non-contact phase, the trajectory of the foot's swing point must follow the normal vector of the ground. And track the reference trajectory v * (t), when in the non-contact term, the contact force is zero:

[0070]

[0071] f ci =0

[0072] When the foot touches the ground, the contact force must satisfy the friction cone constraint:

[0073]

[0074] Where μ is the ground friction coefficient. and These represent the components of the ground contact force in the three directions.

[0075] After optimization by a nonlinear model predictive control (NMPC), an optimized state variable trajectory is obtained. and the trajectory of input variables

[0076] In one possible embodiment, Whole Body Control (WBC) can be used to track the optimized trajectory described above. The WBC controller reallocates contact forces, generalized acceleration, and joint torques through hierarchical quadratic programming to track the optimized trajectory while satisfying dynamic constraints and priority tasks. The control variables of the WBC are as follows:

[0077]

[0078] A priority-based task allocation method is used, dividing tasks into dynamic constraint tasks and target tracking tasks, with dynamic constraint tasks being prioritized. Dynamic constraint tasks include floating base dynamic constraints, joint moment limitation constraints, contact foot non-slip constraints, and friction cone constraints.

[0079]

[0080] Target tracking tasks include swing foot tracking, floating base, robotic arm joint commands, and contact foot force tracking.

[0081]

[0082] J ci This represents the contact point c. i The floating base of the bicorn at that location, and These represent the stiffness and damping coefficients, respectively, for tracking the foot's swing trajectory. and These represent the optimized foot joint angles. Calculate the desired position and velocity of the foot. and These represent the actual position and speed of the foot, respectively. and The stiffness and damping coefficients represent the end effector's trajectory tracking. and This represents the current joint angle and joint speed of the robotic arm.

[0083] By planning the ZMP trajectory of a legged robotic arm and adding the ZMP trajectory tracking loss function to the robot's nonlinear model predictive controller, optimized state and input trajectories are obtained. Finally, a whole-body motion controller based on hierarchical quadratic optimization is used to track the optimized trajectory, thus realizing the task of bipedal walking of the legged robotic arm.

[0084] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A gait switching control method for a legged robotic arm robot, characterized in that, Includes the following steps: S1 calculates the position of the zero-moment point of the current gait and the gait to be switched of the legged robotic arm robot, respectively; Based on the position and linear velocity of the zero-moment point of the current gait and the gait to be switched, as well as the time from the current gait to the gait to be switched, S2 plans the motion trajectory of the zero-moment point of the legged robotic arm robot from the current gait to the gait to be switched. This motion trajectory serves as the reference trajectory of the zero-moment point. S3 uses a predictive control algorithm to control the legged robotic arm robot to move according to the zero-moment point reference trajectory. It calculates the tracking error between the actual motion trajectory and the zero-moment point reference trajectory in real time, uses the tracking error to calculate the loss function of the predictive control algorithm, and adjusts the control parameters of the predictive control algorithm in real time according to the loss function. At the same time, it outputs the state trajectory and the input trajectory in real time, thereby realizing the gait switching of the legged robotic arm robot. The predictive control algorithm takes the zero-moment point reference trajectory as input and the state trajectory and the input trajectory as output.

2. The gait switching control method for a legged robotic arm robot as described in claim 1, characterized in that, The loss function expression obtained in step S3 is: Where x and u represent the state trajectory and input trajectory, respectively, and P(x,t) is the position of the robot's zero-torque point at the current time t. ref For the zero-moment point reference trajectory, the positive definite matrix Q and the positive semi-definite matrix R are the weight matrices for tracking the reference state trajectory and the input trajectory, respectively. P It tracks the reference trajectory P(t) at the zero torque point. ref The weight matrix.

3. The gait switching control method for a legged robotic arm robot as described in claim 1, characterized in that, The constraints of the predictive control algorithm in step S3 include floating base dynamic constraints, no-slip constraints, contact force constraints, and friction cone constraints.

4. The gait switching control method for a legged robotic arm robot as described in claim 3, characterized in that, The dynamic constraint expression for the floating base is: The expression for the no-slip constraint is: The expression for the contact force constraint is: f ci ＝0 The expression for the friction cone constraint is: in, The linear and angular momentum at the center of mass of the legged robotic arm represent the linear and angular momentum. Representing generalized coordinates, including the six-dimensional coordinates q of the robot's base coordinate system in the inertial frame. base q leg q represents the leg joint angles of a quadruped robot. arm A represents the joint angle of the robotic arm. b and A j Represents the center-of-mass momentum matrix A com The corresponding parts of the fuselage and the corresponding parts of the joints, f ci and These represent the vector pointing from the center of mass to the contact point and the three-dimensional contact force exerted by the ground on the robot, respectively. Represents the speed between the foot tip and the ground. The normal vector representing the ground, v * (t) represents the reference velocity, and μ represents the ground friction coefficient. and These represent the components of the ground contact force in the three directions.

5. The gait switching control method for a legged robotic arm robot as described in claim 1, characterized in that, In step S3, the state trajectory includes the linear momentum and angular momentum at the center of mass of the legged robotic arm in the current state, the joint angles of the legs, and the joint angles of the robotic arm; the input trajectory includes the three-dimensional contact force of the ground on the legged robotic arm, the joint velocities of the legs, and the joint velocities of the robotic arm.

6. The gait switching control method for a legged robotic arm robot as described in claim 1, characterized in that, The method for calculating the location of the zero torque point in step S1 is as follows: Where, x cog y cog and z cog p represents the coordinates of the robot's center of gravity in various directions in the inertial frame. x p y and p z It is a symbolic representation of the three-dimensional coordinates of the zero-torque point.

7. The gait switching control method for a legged robotic arm robot as described in claim 1, characterized in that, The specific implementation method of step S2 is to use Bézier curve fitting to generate a zero-moment point reference trajectory, which is as follows: P(τ)=(1-τ) 3 A+3(1-t) 2 τC+3(1-τ)τ 2 D+t 3 B t=τ·T traj In this context, control points A and B represent the initial and final two-dimensional coordinates of the zero-torque point, while control points C and D are located along the initial velocity direction of the zero-torque point, respectively. and final velocity direction and Representing the initial and final velocity values, the parameter τ∈[0,1] represents the parameters of the entire process from the initial point P1 to the final point P2, and t is the time scale. traj This represents the total duration of the zero-torque point trajectory from the initial point P1 to the final point P2.

8. A gait switching control system for a legged robotic arm, characterized in that, For implementing the gait switching control method according to any one of claims 1-6, the gait switching control system comprises: The calculation module is used to calculate the zero-torque point position of the legged robotic arm robot; Zero-moment point planning module, which is used to plan the zero-moment point reference trajectory of the legged robotic arm robot from the current gait to the gait to be switched; The predictive control module is used to construct a predictive control algorithm with a zero-torque point reference trajectory as input and a state trajectory and an input trajectory as output. The predictive control algorithm is used to control the movement of the legged robotic arm robot and outputs the state trajectory and input trajectory in real time to control the gait switching of the legged robotic arm robot.

Images