A Collision-Free Framework for Navigation of Nonholonomic Vehicle Systems

The task of highly effective navigation is certainly one of the critical technologies of autonomous vehicles that have been widely investigated by researchers. In this paper, a navigation framework with path planning and tracking is presented, where a feasible and collision-free trajectory is generated for nonholonomic vehicle systems. In terms of path planning, an LQR-RRT* algorithm is applied to plan a path without collision. It is an optimal sampling-based planning algorithm with linear quadratic regulation-based heuristics, and able to tackle some rather complicated or under-actuated dynamic systems. In order to improve the feasibility of the planned path and compensate for the deterioration of optimality due to sampling times, the approach is taken where a cubic spline path is additionally generated. The path tracking problem is formulated as a constrained optimization problem for the nonholonomic vehicle system, and then the required steering angle and acceleration is computed to avoid potential collisions. The results of the simulation experiments clearly demonstrate the effectiveness of the proposed navigation framework.