Research On Online Motion Planning Of Redundant Manipulators With Environmental Constraints

In the operation and tasks of multifunctional maintenance manipulator with environmental constraints,online motion planning can reasonably decompose the tasks according to the limited sensing resources,and make the robot automatically make safe,orderly and stable actions.At present,some limitations still exist in the theoretical research on online motion planning,such as slow convergence rate,model-based obstacle avoidance,high complexity of joint trajectory planning,which limits the practical application of motion planning theory in multifunctional maintenance arm to a certain extent.In addition,the existing robot system owns open kinematics interface that is conducive to the integration of advanced intelligent algorithm to realize robot automation and intelligence.Therefore,the research on online motion planning method is of great practical significance for the promotion of motion planning theory of multifunctional maintenance manipulator and the further development of online motion planning theory.In this paper,the basic theory of motion planning for multi-functional maintenance arm is explored.Based on kinematics,the motion planning of redundant planar arm in dynamic environment,the motion planning of redundant threedimensional configuration manipulator,the angular acceleration planning based on joint space and the redundant multi arm motion planning are studied.The study on online motion planning in the dynamic environment for redundant planar manipulator is presented firstly.This paper designs an adaptive virtual controller based on the pose error to improve the convergence speed of motion planning.The motion observation and path prediction are integrated into the motion planning,and the motion paths of the object and obstacle are fitted by spline filter and are predicted by the defined second-order polynomial.Similarity evaluation verifies that the proposed method is effective.The criterion of judging the shortest path is defined to ensure the accessibility of the end-effector motion path.At the same time,the model-based local rotation coordinate method is proposed to avoid the obstacle for the end-effector,which overcomes the disadvantages of traditional obstacle avoidance method such as small range obstacle avoidance and easy to fall into local minimum value.Finally,the algorithm of online motion planning is formed in the dynamic environment for redundant planar manipulator.In the research on online motion planning of redundant three-dimensional configuration manipulator,this paper uses the single neuron PID to model the position and attitude error of the end-effector,and utilizes the unsupervised principal component analysis to adjust the neuron weight coefficient adaptively.Through the competition between the neuron weights,the neural model achieves a stable state,and realizes the smooth and fast convergence,avoiding the problem of large initial gain and instability in traditional methods.The motion state of the robot arm is described by energy.The energy describing the motion around obstacles and the energy describing the tendency towards object motion are defined for the end-effector obstacle avoidance.The energy maintaining a safe distance from the obstacle and the energy moving away from the obstacle are defined for the arms of the robot.A continuous smooth transformation between different energies is designed.The differentiable Sfunction guarantees the safety and collision-free of the robot arm in the environment of dynamic obstacles.The attitude adjustment method for the end-effector is proposed to avoid the failure of the manipulator.The adaptive learning algorithm and obstacle avoidance method based on energy concept are integrated to form an online motion planning method for redundant three-dimensional manipulator.On the aspect of the angular acceleration planning in joint space,this paper defines Cartesian space pose error model,establishes joint space mapping relation based on generalized inverse kinematics,and proves the vertical orthogonal and linear independent relation between null-space vector of redundant manipulator and main task vector.The joint velocity formed by inverse kinematics mapping is regarded as system error,and then the joint angular acceleration model is derived,which simplifies the highly nonlinear complex term of the system,and realizes the transformation from planning problem to control problem.Based on the theory of high-order sliding mode control,the hyperbolic tangent supper twisting control algorithm,which is used as the control input,is designed to suppress the system disturbance,ensure the system convergence,and avoid the joint vibration caused by the traditional high-order sliding mode in real time.The gain function with the null-space velocity and system error is designed to prevent integral saturation.The real-time control of joint angular acceleration is realized by combining the real-time integration of control input.The Lyapunov function is designed,the stability of the system and the range of parameters are derived and analyzed in detail.The effects of the parameters on the convergence and the motion path of the end effector are discussed.The performances of the proposed method and the traditional method in null-space obstacle avoidance of redundant manipulator are compared.Then,in the research on the online motion planning method of redundant multiarm robot,this paper studies the planning problem of redundant multi-arm robot with common motion physical coupling,and divides the task of multi-arm robot into independent operation task and cooperative operation task.Based on the proposed energy conversion strategy-based obstacle avoidance method,the collision free independent tracking task of the multi-arm robot is realized.In the multi-arm cooperation task,the sub-base method is proposed to divide the robot configuration.The inverse kinematics mapping of the multi-arm robot is realized based on the leastsquare Jacobian inverse to ensure the synchronization of the motion state of the multiarm robot.Based on the inner star learning rules and the idea of self-organizing competitive neural network model,a motion planning method for multi-arm coordination task is designed,which improves the synchronization and coordination of multi arm motion.Based on Lyapunov theory and the principle of inner star learning rules,the stability and convergence of the proposed motion planning method based on self-organizing competitive neural network are analyzed.The application of the proposed planning method to redundant dual-arm and three-arm robots,as well as the applicability and synchronization of the multi-manipulator with fixed common connection are discussed.Finally,the experimental platform of 7-degree-of-freedom redundant robot manipulator guided by monocular vision and 13-degree-of-freedom dual-arm robot based on depth vision are designed and built.The feasibility and effectiveness of the proposed redundant planar manipulator motion planning,energy conversion obstacle avoidance strategy-based motion planning,joint angular acceleration planning and multi-arm motion planning based on self-organizing competitive neural network is verified.The real-time moving object tracking experiments verify that the proposed method has the large-scale turning and obstacle avoidance,fast convergence of virtual controller,energy-based model-free obstacle avoidance,and fast smoothness of single neuron PID autonomous programming based on principal component analysis learning rule.The experiment of joint angular acceleration planning verifies the smoothness of joint angle and angular velocity trajectory and the convergence of error in the process of tracking moving object in dynamic obstacle environment.The feasibility and synchronization of the dual-arm motion planning method are verified by six kinds coordinated operation experiments.