In this paper, we present a novel approach within the sequentialframework to develop practical motion planners for many degrees offreedom (DOF) arms. In this approach, each of the sub-problem is solvedby using numerical potential fields defined over bitmap-basedrepresentations of the 2-dimensional sub-spaces. Furthermore, anefficient backtracking mechanism based on a novel notion of virtualforbidden regions in these 2-dimensional subspaces is presented. Thisnovel approach leads to much more efficient and robust motion plannersthan a previously reported visibility graph (in the 2-dimensionalsubspaces) based implementation. We have conducted extensive experimentsfor planar arms with up to 8-DOF among randomly placed obstacles.Although it is not complete, the planner never failed for the examplesin hundreds of simulations, and very small backtracking levels wereneeded. We have implemented the planner for 3-dimensional workspaces andan illustrative example for a 7-DOF manipulator shows the promise of ourapproach
展开▼
