Evolving a multiobjective obstacle avoidance skill of a seven-link manipulator subject to constraints

Abstract

A remarkable part of the skills of human experts rendering services in hazardous and dynamic factory environments is their ability to continue to perform a job while reacting to adapt their posture to sudden changes in the environment such as upcoming obstacles. Possession of these typical human skills that can be defined in a set of linguistic objectives will help to build flexible and more dexterous robot manipulators for factory automation systems that will meet the challenges of the growing complexity of the modern automated manufacturing environments. In this paper a novel soft-computing-based method is proposed for redundancy resolution of industrial robot manipulators to perform dexterous obstacle avoidance motions subject to the working constraints of the end effector. The new approach is based on evolving redundant rows of the Jacobian matrix of the manipulator and intelligent manipulation of user-defined kinematic functions for redundancy resolution. An evolutionary approach has been adopted to enhance the search efficiency in the objective space of a linguistic multi-objective problem (MOP) for the task of avoiding an obstacle moving towards the manipulator while keeping the end-effector position and orientation vector static. The objective functions of the MOP appear in the weighted sum of objectives in a gradually growing manner. The basic natural phenomena in the process of complex skill acquisition by human beings are reflected in the proposed approach. The effectiveness of the proposed method is demonstrated through experiments carried out using an industrial seven-link manipulator called PA-10, manufactured by the Mitsubishi Heavy Industries Ltd.