Empirical Evaluation of Theta∗-RRT and GRIPS Algorithms

Motion planning is a fundamental task for wheeled mobile robots. This task becomes challenging when the kinematic constraints of the robot (differential-drive, car-like, etc.) are to be taken into account. In this work we analyze and empirically compare two promising approaches to construct kinematically-feasible trajectories for differential drive robots – Theta*-RRT and GRIPS. Both of these approaches utilize geometric path planning but differ in the way how it is done. Theta*-RRT relies on sampling-based planning biased towards the geometric path. GRIPS modifies the path and tries to connect its elements with the steering function that respects kinematic constraints. We evaluate both approaches in simulation and on the real robot highlighting their pros and cons. Our evaluation shows that there is no universal winner and we provide suggestions on when to use specific method.