Hot-start based Fast Speed Planning for Eco-Driving of Intelligent Vehicles

This paper proposed a hot-start based fast optimal speed planning method for intelligent vehicles through multiple intersections. The purpose is to improve the real-time performance of current methods. The optimal control problem is formulated into a nonlinear optimization problem. A hierarchical architecture is constructed, where the upper layer utilizes dynamic programming(DP) to provide constraints and an initial guess of the nonlinear optimization at the lower layer. At the upper layer, based on cruise speed economy analysis, a weighted orientation graph problem can be established. The optimal solution of the graph is obtained via DP, which determines the optimal passing green phase at each intersection. At the lower layer, the interior point method for nonlinear optimization is adopted. We use the DP solution as an initial guess of the interior point method, which is a hot-start mechanism for fast convergence in the process of optimization iteration. Simulation results demonstrate that the proposed hot-start mechanism reduces the computational time by up to 68.4% compared with a rule-based initial guess method. The proposed global speed planning method is able to be implemented in real-time while maintaining the same speed profile optimality.