Balancing state-space coverage in planning with dynamics

Sampling-based kinodynamic planners, such as the popular RRT algorithm, have been proposed as promising solutions to planning for systems with dynamics. Nevertheless, complex systems often raise significant challenges. In particular, the state-space exploration of sampling-based tree planners can be heavily biased towards a specific direction due to the presence of dynamics and underactuation. The premise of this paper is that it is possible to use statistical tools to learn quickly the effects of the constraints in the algorithm's state-space exploration during a training session. Then during the online operation of the algorithm, this information can be utilized so as to counter the undesirable bias due to the dynamics by appropriately adapting the control propagation step. The resulting method achieves a more balanced exploration of the state-space, resulting in faster solutions to planning challenges. The paper provides proof of concept experiments comparing against and improving upon the standard RRT using MATLAB simulations for (a) swinging up different versions of a 3-link Acrobot system with dynamics and (b) a second-order car-like system with significant drift.