Obstacle Avoidance for Kinematically Redundant Manipulators Based on Recurrent Neural Networks

摘要

As the development of automation industry, robot manipulators are needed to work in more and more complex and dynamic environments. An important issue to be considered is how to avoid static or moving obstacles in the workspace. Kinematically redundant manipulators are those having more degrees of freedom than required to perform end-effector moving tasks in a given workspace. Being dexterous and flexible, they have been used for avoiding obstacles or singularity, and optimizing various performance criteria in addition to tracking desired end-effector trajectories. Of those versatile applications, obstacle avoidance is extremely important for successful motion control in the presence of obstacles.