Planning Robot Motions to Satisfy Linear Temporal Logic, Geometric, and Differential Constraints

This paper shows how to effectively compute collisionfree and dynamically-feasible robot motion trajectories that satisfy task specifications given by Linear Temporal Logic (LTL). The proposed approach combines sampling-based motion planning over the continuous state space with discrete search over both the LTL task representation and a workspace decomposition. In distinction from related work, the proposed approach samples the discrete space to shorten the length of the discrete plans and to more effectively guide motion planning in the continuous state space. Experimental results on various scenes, LTL specifications, and a snake-like robot model with nonlinear dynamics and numerous degrees-of-freedom (DOFs) show significant computational speedups over related work.