Research on inverse kinematic of concentric tube robot based on reinforcement learning
Concentric tube robot(CTR)is a type of continuum robot with the characteristics of small size,scalability,and high curvature,which is suitable for the field of minimally invasive medicine.However,due to the complexity of its kinematic model,implementing effective inverse kinematics control is a challenging task.This paper intro-duces the forward kinematic model of CTR to realize the mapping from the control input to the Cartesian position.An inverse kinematic control method based on reinforcement learning(RL)is proposed.Discrete reinforcement learning training actions are designed according to the actual drive system,and training rewards are designed in combination with the actual drive limit.Moreover,the asynchronously advantage actor-critic(A3C)learning strate-gy is used to calculate CTR control input that is more suitable for practical applications.It provides a new method for the inverse kinematics control of CTR.In addition,the CTR control platform is designed and built to verify the effectiveness and correctness of the proposed inverse kinematics control method.In trajectory tracking experiments,the CTR can track the trajectory with an average tracking error of 1.462±0.483 mm.
万方数据
维普
