Semi-lazy probabilistic roadmap: a parameter-tuned, resilient and robust path planning method for manipulator robots

An indispensable feature of a modern intelligent robot is its capability to plan short and safe motions in the presence of obstacles in its workspace, which is highly important for industrial manipulators in charge of automatic picking and placing, welding, painting, etc. On the other hand, collision-free motion planning of serial manipulators becomes exponentially hard with the increase of number of joints, and so efficient methods like sampling-based ones are vastly used for most real-world problems. In this paper, we propose a new variation of sampling-based methods called semi-lazy probabilistic roadmap (SLPRM) for motion planning of industrial manipulators, which benefits from the advantages of the basic probabilistic roadmap (PRM) and lazy-PRM (LPRM) methods. Unlike the exhaustive and zero collision-checking policies implemented respectively in PRM and LPRM, the SLPRM collision-checks random configurations for only m terminal links (i.e., from end-effector backwards) of the manipulator in the roadmap construction phase. As a result, on one hand, the roadmap construction time reduces compared with PRM due to less collision checks, and on the other hand, query times decrease compared with LPRM due to a better quality of the initial roadmap. A central decision in SLPRM is to properly determine the value of m, which has a direct effect on its speed. For this purpose, a new parameter tuning approach based on a combination of Shannon's Entropy and VIKOR methods is implemented to determine the best values for m and all other parameters of the algorithm. The proposed method has been tested and implemented in simulated and real workspace scenarios for an RV-E3J Mitsubishi industrial manipulator robot, and the results showed that the mean planning time of the SLPRM was shorter compared with that of the PRM and LPRM. To make the algorithm resilient and robust to internal faults and environmental variations such as positional errors, joint failures, and obstacle displacements, we have also proposed the resilient and robust SLPRM, which through concentrated sampling and roadmap-amending procedures, can handle unexpected failures and changes.