Poli-RRT*: Optimal RRT-based planning for constrained and feedback linearisable vehicle dynamics

摘要

This paper proposes a Rapidly exploring Random Trees planning strategy (Poli-RRT*) that computes optimal trajectories in presence of vehicle constraints (e.g., differential and actuation constraints) without approximating the nonlinear dynamics, but relying on exact linearisation. In this way, the optimal control problem that is introduced to determine the trajectories extending the tree can be expressed as a quadratic program and efficiently solved. Poli-RRT* is formulated and tested via simulation on a unicycle-like model of a vehicle subject to actuation constraints. Notably, the approach can be applied to any other feedback linearisable vehicle model, subject to different types of constraints.