Electric Vehicle Routing Problem for Emergency Power Supply: Towards Telecom Base Station Relief

TL;DR

This work addresses maintaining telecom base station power during outages by routing electric vehicles (EVs) to directly supply stations, formulating EVRP-EPS as a new variant of the EVRP. The authors propose a solver that pairs a rule-based vehicle selector with an RL-based node selector, implemented via a centralized two-tower Transformer policy, and trained with REINFORCE. Empirical results on synthetic and real datasets show the approach achieves superior objective values and faster runtimes than baselines, with demonstrated scalability and generalization to unseen settings. The work advances emergency-power routing by integrating continuous battery dynamics, action-cycle constraints, and efficient route generation suitable for rapid deployment in disaster scenarios.

Abstract

As a telecom provider, our company has a critical mission to maintain telecom services even during power outages. To accomplish the mission, it is essential to maintain the power of the telecom base stations. Here we consider a solution where electric vehicles (EVs) directly supply power to base stations by traveling to their locations. The goal is to find EV routes that minimize both the total travel distance of all EVs and the number of downed base stations. In this paper, we formulate this routing problem as a new variant of the Electric Vehicle Routing Problem (EVRP) and propose a solver that combines a rule-based vehicle selector and a reinforcement learning (RL)-based node selector. The rule of the vehicle selector ensures the exact environmental states when the selected EV starts to move. In addition, the node selection by the RL model enables fast route generation, which is critical in emergencies. We evaluate our solver on both synthetic datasets and real datasets. The results show that our solver outperforms baselines in terms of the objective value and computation time. Moreover, we analyze the generalization and scalability of our solver, demonstrating the capability toward unseen settings and large-scale problems. Check also our project page: https://ntt-dkiku.github.io/rl-evrpeps.

Figures (5)

• Figure 1: An illustration of telecom base station relief. EVs directly supply power to base stations by traveling to their locations. Each EV returns to a charge station located outside the blackout area before the EV power runs out.
• Figure 2: An illustration of the EV action cycle: move$\to$prepare$\to$(dis-) charge$\to$clean-up$\to$move$\to$$\cdots$
• Figure 3: An overview of our solver. It first selects the EV that will be the earliest available for the next move. Then, it determines the next destination of the selected EV with a centralized stochastic policy, which is parameterized by a two-tower Transformer.
• Figure 4: Four visual results for each real dataset: The time variation in the number of downed base stations (1st row), the time variation in base station/EV batteries (2nd-3rd rows), and the route network (4th row). Each column corresponds to each of the real datasets and includes those corresponding results.
• Figure 5: Results of scalability test (top row) and generalization test (bottom row). The points represent average values. The filled area indicates the standard deviation. The dotted line represents the linear increase from the starting point.