Multi-objective autonomous eco-driving strategy: A pathway to future green mobility
Tong He, Liang Chu, Zheng Chen, Y Q Liu, Yuanjian Zhang, Jincheng Hu
Abstract
With the wide popularity of electric vehicles in the market and advancements in autonomous driving technology, intelligent electric vehicles (iEVs) equipped with comprehensive eco-driving capabilities are expected to play a pivotal role in energy conservation and emission reduction of future mobility. This paper proposes an intelligent eco-driving strategy (IEDS) to address the safety and eco-driving concerns with the parallel hybrid electric vehicle (PHEV). The IEDS is a data-driven autonomous driving solution to effectively control vehicle motion and energy management, developed based on refined deep reinforcement learning (DRL) algorithms, integrating safety and efficiency knowledge in autonomous driving through a multi-head deep Q network (DQN) with elaborate rewards for potentially dangerous collisions and fuel consumption. In the case studies, the simulations show that the IEDS is able to achieve excellent energy-saving performance through stable and safe driving manners. Compared with the baselines, its obstacle avoidance and energy-saving performance are 2.10% and 5.83% ahead, achieving 97.07% of the optimal energy management result. • Advanced IEDS boosts safety, efficiency in autonomous driving with PHEVs. • Multi-head DQN enhances eco-driving through effective energy and collision control. • IEDS achieves 2.10% better obstacle avoidance and 5.83% more energy savings. • Overcomes adaptive optimization challenges, reaching 97.07% of optimal performance.