Vehicle dynamics oriented motion planning methodology for autonomous vehicles using iterative linear quadratic regulator

To address the degradation in trajectory smoothness and motion stability of autonomous vehicles (AVs) under extreme conditions, this paper proposes a vehicle dynamics-oriented motion planning methodology utilizing the iterative Linear Quadratic Regulator (iLQR). Within the planning methodology, the initial geometric trajectory is calculated based on polynomial spline curves in the Frenet curvilinear coordinate system, where Bezier curve constraints are linearly mapped to ensure continuous obstacle avoidance performance, preventing constraint violations between trajectory points. Building on this geometric foundation, an optimization problem incorporating vehicle dynamics is formulated for high-performance motion planning, in which the nonlinear vehicle characteristics are explicitly integrated and comprehensive performance indicators have been designed considering motion stability and tire adhesion performance. Furthermore, an iterative Linear Quadratic Regulator method is employed to efficiently address the optimization problem with complex constraints. This approach balances model fidelity and computational efficiency, generating feasible trajectories that adhere to the vehicle inherent dynamic characteristics. The simulation results in the overtaking scenario, along with the actual vehicle experiments, demonstrate that the proposed methodology can significantly enhance the geometrical smoothness and motion stability compared to the existing methods, which guarantees effectiveness and capability in the nonlinear driving scenario. • Optimization-based polynomial spline planning in the Frenet frame guarantees strict adherence to geometric and obstacle constraints via linear Bezier constraint mapping , ensuring continuous feasibility and eliminating inter-point violations. • ciLQR optimization with nonlinear vehicle dynamics integrates comprehensive stability indicators (tire adhesion, motion stability) to refine trajectories efficiently, balancing computational speed and dynamic model accuracy. • Validation through high-fidelity simulations and vehicle experiments demonstrates superior trajectory smoothness and motion stability, confirming real-time feasibility of the algorithm based on high-complexity dynamics in extreme conditions.