Motion Planning Algorithms for General Closed-Chain Mechanisms

Abstract

A robot essentially acts by moving in a physical world. The motion planning capability is thus a fundamental issue for the autonomy of the system, and represents a very active area of research in Robotics. Furthermore, the interest in motion planning techniques goes beyond robotic applications. Currently, these techniques are applied in other very different domains such as : CAD/CAM, industrial logistics, graphic animation, or computational Biology. Complex articulated mechanisms containing closed kinematic chains appear in all these domains. This thesis treats motion planning for such systems. The first part contains our theoretical and technical work. We introduce samplingbased planners into a general formulation of the motion planning problem in presence of kinematic closure constraints. The algorithmic tools that we propose allow an efficient application of these techniques to complex closed-chain mechanisms. The second part deals with the different fields of application that we have investigated. In Robotics, our approach has been applied to motion synthesis of parallel robots, coordinated manipulation and manipulation planning. Finally, we discuss an original application to Structural Biology for the conformational analysis of protein loops. The results of our experiments prove the efficacy and the generality of the approach.