Linear Temporal Logic Guided Motion Planning for Robots with Hybrid Locomotion Dynamics
In this thesis, we consider the problem of motion planning for robots with hybrid locomotion dynamics to achieve complex tasks specified by linear temporal logic (LTL) formulas. We first present a novel planning technique for non-periodic bipedal locomotion tasks specified by co-safe LTL formulas in rough cluttered terrains. Our planning approach is based on a discrete set of motion primitives for the center of mass (CoM) of a general bipedal robot model. A deterministic shortest path problem is solved over the deterministic finite-state automaton (DFA) of the temporal logic task specification, composed with the graph of CoM keyframe states generated by the motion primitives. We then address the infinite-horizon planning problem presented by the use of full LTL semantics for task specification. Here, we use a discrete set of motion primitives for the continuous state of a robot model with hybrid locomotion dynamics. We divide the problem of finding a trajectory satisfying the task into finding an accepting cycle and the acyclic trajectory connecting to it from the initial state. To find these trajectories we propose an anytime graph-search algorithm that provides guarantees on the maximum cost of the total trajectory. The approaches presented in this thesis enable motion planning for robots with hybrid dynamics to achieve complex tasks specified by linear temporal logic formulas.