Energy-Efficient Capacitated Electric Vehicle Routing Problem Solved with Token-Specific Deep Reinforcement Learning

Researchers from Xi'an Jiaotong University have made a groundbreaking discovery in the field of electric vehicle routing problems (EVRPs). The team, led by Dapeng Yan, has developed a novel approach called token-specific deep reinforcement learning (TS-DRL) to tackle the complex challenges of EV routing. This innovative method incorporates real-world factors such as road conditions, driving dynamics, and battery efficiency to minimize energy consumption.

Key Takeaways:

• The existing approaches to EVRPs primarily focus on minimizing travel distance, neglecting critical factors that significantly influence energy consumption.
• The researchers propose an advanced energy consumption model that incorporates capacity limits, driving resistance, battery efficiency, road characteristics, and vehicle dynamics to capture the multifaceted challenges of EV routing.
• The proposed TS-DRL approach frames the energy-efficient capacitated EVRP (CEVRP) as a customized Markov decision process (MDP) and develops a token-specific framework that integrates real-time EV state information with the active routing sequence.
• Extensive experiments conducted on synthetic and real-world datasets demonstrate the effectiveness of TS-DRL, achieving energy savings of up to 109.86 kWh (i.e., a reduction of 13.79%) in scenarios with 100 customers and eight charging stations.
• The results reveal that TS-DRL consistently outperforms various heuristic and DRL-based methods in tackling more complex EV routing challenges in practical energy and transportation systems.
• The research highlights the potential of TS-DRL to tackle more complex EV routing challenges and provides a framework for future research in this area.

Statistics:

• Up to 109.86 kWh energy savings achieved by TS-DRL in scenarios with 100 customers and eight charging stations (a reduction of 13.79%).
• TS-DRL consistently outperforms various heuristic and DRL-based methods in tackling more complex EV routing challenges.

Sources:

• Yan, D., Guan, Q., Cao, H., Tan, J., Jia, L., & Chen, B. (2025). Token-specific Deep Reinforcement Learning for Energy-efficient Capacitated Electric Vehicle Routing Problems. Applied Energy, 396.
• Xi'an Jiaotong University, School of Electrical Engineering.
• National Key Research & Development Program of China.
• National Natural Science Foundation of China (NSFC).