Energy Optimized Driving Strategy Using Machine Learning

Abstract

This project focuses on developing an energy-optimized adaptive cruise control (ACC) model for battery electric vehicles (BEVs) using the Deep Deterministic Policy Gradient (DDPG) algorithm. The study explores the potential of DDPG in creating an ACC system that maximizes energy efficiency while considering battery life. Battery modeling and degradation models are incorporated to evaluate the performance of the developed model. A comparison with an available Model Predictive Control (MPC) controller demonstrates improvements in capacity loss, energy efficiency, and reduction in cell temperature. However, challenges arise in striking a balance between reducing velocity and distance errors while minimizing current and energy consumption. This project provides a foundation for enhancing energy efficiency and battery life in ACC systems, but further refinement is necessary to ensure suitability for real-world applications. Limitations of the project include a loosened distance constraint and simplified environment and vehicle modeling. Future work involves parameter tuning, refining the reward function, and incorporating more realistic factors.