This paper has presented a novel motion planning technique
designed for modular robots. The proposed method, called RRTMP,
is based on a sampling-based principle that solves the motion
planning problem by random sampling of the configuration space
of the robot. Due to the high number of actuators, it became
intractable to examine all possible combinations of control inputs
of the modular robot. We therefore suggest using local motion
primitives. This allows the RRT-MP to effectively explore the
configuration space. These motion primitives should provide
only short motion in the vicinity of the robot and they can be
implemented e.g. using CPGs.