Guided Motion Planning for Modular Robots

Modular robots consist of many modules that can be connected into various structures. This allows modular robots to adapt their shape according to a given task and environment. To visit a desired place in an environment, motion planning is required. To generate feasible plans for these robots with many degrees of freedom and many kinematic and dynamic constraints, a physical simulation is required to precisely model motion of the robots. As the physical simulation can be time consuming, motion planners need to be extended to decrease the number of simulation queries. In this paper, we propose a novel motion planning for modular robots based on guided sampling of configuration space. The guided sampling utilizes a precomputed simple path in the environment, along which the configuration space of the robot is sampled with higher probability. This approach significantly decreases the number of calls of physical simulations which increases the speed of the planning process.