Smart and efficient EV charging navigation scheme in vehicular edge computing networks

Haoyu Li, Jiming Chen, Chao Yang, Xin Chen, Le Chang, Jiabei LiuJ. Cloud Comput 2023 · DOI

Introduction

As smart cities evolve, Electric Vehicles (EVs) are transitioning from simple energy consumers to active participants in digital infrastructure. A recent study published in the Journal of Cloud Computing proposes a novel navigation scheme that leverages EVs as both power loads and mobile edge computing nodes. By synchronizing charging schedules with the computational needs of the network, the system optimizes travel efficiency while potentially generating revenue for vehicle owners.

The Challenge of Long-Distance EV Travel

Long-distance EV travel is currently hindered by three primary factors: unpredictable traffic congestion, fluctuating electricity prices, and the significant "dead time" spent at Fast Charging Stations (FCSs). While existing navigation systems focus on the shortest path or lowest energy cost, they often overlook the computational potential of modern EVs. In a Vehicular Edge Computing Network (VECN), roadside units (RSUs) frequently encounter processing bottlenecks. This research addresses how to turn an EV's downtime during charging into a productive window for the network.

System architecture of the proposed vehicular edge computing network (VECN).

Proposed Methodology: A Two-Stage Approach

The researchers formulated the navigation problem as a Mixed Integer Programming (MIP) challenge, balancing travel time, energy expenditure, and computational rewards. To manage the dynamic nature of urban environments, they implemented a two-stage algorithm:

  1. Stage 1: Global Route Planning: Utilizing the A* algorithm and historical traffic data from a Cloud Computing Center (CCC), the system identifies the optimal path from the starting point to the destination.
  2. Stage 2: Local Decision Making: As the vehicle progresses, a Deep Q-Network (DQN) handles real-time adjustments. This stage determines which specific FCS to utilize and the optimal "staying time."
Deep Q-Network (DQN) framework for local charging and task offloading decisions.

The inclusion of "staying time" as a variable is a key innovation. An EV might remain at a station beyond its full charge if the road ahead is congested or if the rewards for sharing its idle computing power are particularly high at that moment.

Simulation and Experimental Results

The study's effectiveness was validated through a custom-built simulation environment featuring a grid-based road network with stochastic congestion. The vehicle model was based on standard market EVs, and economic factors included real-world Time-of-Use (TOU) electricity pricing.

The proposed scheme demonstrated several significant advantages:

  • Economic Viability: In multiple scenarios, the rewards earned from task offloading were sufficient to offset charging and travel costs, resulting in a negative total system cost—essentially, the EV "earned" money during the trip.
  • Resilience to Congestion: By treating the EV as a "flexible load," the algorithm successfully navigated around traffic peaks. Strategic waiting at an FCS allowed vehicles to avoid energy-intensive stop-and-go traffic while continuing to earn rewards.
  • Algorithmic Performance: The DQN-based approach significantly outperformed standard Q-learning and "SoC-only" baselines, maintaining stability and efficiency across routes ranging from 600km to 1200km.

Limitations and Future Directions

While promising, the study identifies several areas for further development. The current model treats computational tasks as generic, whereas real-world applications have varying hardware requirements. Additionally, the security and privacy implications of sharing onboard compute resources with a public network remain to be explored. Future work may involve integrating blockchain for secure transactions or employing multi-agent reinforcement learning to model interactions between fleets of competing EVs.

Summary updated

BibTeX
@article{li2023smart,
  author = {Haoyu Li and Jiming Chen and Chao Yang and Xin Chen and Le Chang and Jiabei Liu},
  title = {{Smart and efficient EV charging navigation scheme in vehicular edge computing networks}},
  journal = {J. Cloud Comput.},
  year = {2023},
  doi = {10.1186/s13677-023-00547-y},
}
Figure