A new Probabilistic Path Planning Algorithm for (Dis)assembly Tasks

In this paper, a new probabilistic path planning algorithm is described. The algorithm has been developed for assembly planning purposes, but however it can also be used in similar scenarios. Most assembly planning algorithms apply the so called assembly-by-disassembly strategy, therewith planning starts from the goal position of parts and tries to remove single parts or group of parts. Such problems are characterized by the appearance of many narrow passages. Thus, we have developed an probabilistic algorithm able to ??nd paths even if many of such passages exist. The idea of our path planner emanates from the particle In each iteration, it propagates new samples, discards bad evaluated samples and assigns higher weights to good examples. The evaluation functions propagates the samples to explore free space as well as to condensate on the border of obstacles, which leads samples to pass narrow passages. We have evaluated our path planning algorithm with different examples and compared it to the well-known RRT and PRMplanners. We could achieve good execution times for realistic industrial assembly tasks.