A reinforcement motion planning strategy for redundant robot arms based on hierarchical clustering and k-nearest-neighbors

Redundant robot arms refer to those robotic manipulators having more degrees-of-freedom than required for a given task, and human arms are inherently redundant. Due to the redundancy, this type of manipulators is very dexterous and agile to accomplish many challenging tasks, such as catching objects in flight, avoiding obstacles, etc. However, with the extra degrees-of-freedom, there exists no close form solution for the motion planning or inverse kinematic problem of redundant robot arms. In this paper, a novel strategy combining hierarchical clustering and k-nearest-neighbors (KNN) is proposed to solve this problem. Random sampling based on Gaussian distribution is applied to generate training set for the learning algorithms. K-means is then used to conduct the hierarchical clustering. And KNN is applied to calculate the output joint angles for the corresponding input end-effector states during test period. Simulation results in MATLAB performed on a five degrees-of-freedom planar robot have demonstrated the high accuracy and computation speed of this method, moreover, this method has also the precious capability of repetitive motion planning.