Experimental Evaluation of a Criteria-Based Obstacle Avoidance Scheme

摘要

Abstract Obstacle avoidance allows manipulators to work in cluttered environments without damaging themselves or their environment. It generally involves collision prevention and configuration optimization to avoid obstacles, with collision prevention being the only choice for non-redundant manipulators. In the case of redundant manipulators, one obstacle avoidance scheme uses null-space options for a given end-effector location, and subsequent ranking of these options to select the ‘best’ one based on input criteria. Under this scheme, four obstacle avoidance criteria have been developed. These four (and other non-obstacle avoidance) criteria along with workspace modeling have been implemented as an extension of the OSCAR C++ software library. This library can be used to develop obstacle avoidance applications involving multiple manipulators operating in complex environments. Experimental evaluation of this scheme is conducted on a 17 degree of freedom dual-arm robotic manipulator, and results are presented.