A Stochastic Particle Filter Energy Optimization Approach for Power-Split Trajectory Planning for Hybrid Electric Autonomous Vehicles

Abstract

In recent years, all major car manufacturers have started to introduce predictive functionalities based on an electronic horizon for the autonomous on-highway operation of their vehicles. Using the Advanced Driver Assistance Systems (ADAS) for anticipatory driving is a fundamental approach to significantly reduce the fuel consumption and pollutant emissions of the internal combustion engines. Today's Adaptive Cruise Control (ACC) systems try to maintain a constant speed selected by the driver without regarding the energy consumption of the vehicle. There is, however, a degree of freedom to apply cruise speed limits without direct driver involvement in order to save propulsion energy in an Autonomous Vehicle (AV). This work presents a novel velocity and energy optimization method for AV hybrid electric vehicles (HEVs) by using a stochastic optimization technique. By applying Particle Filters (PFs) in a routine of Stochastic Dynamic Programming (SDP) to solve the power split efficiently. The sweet-point operation of the powertrain is calculated by probability hypothesis densities along the distance-based prediction horizon. The optimization approach shows a cost trade-off between horizon resolution, length, iterative combinatorial optimality, and computational efficiency. Finally, the approach is applied to a PHEV vehicle model in a real-time ECU in the Worldwide harmonized Light Duty Test Cycle (WLTC).