Decentralized Assignment of Electric Vehicles at Charging Stations Based on Personalized Cost Functions and Distributed Ledger Technologies

Michela Moschella, Pietro Ferraro, Emanuele Crisostomi, Robert ShortenIEEE Internet of Things Journal 2021 · DOI

TL;DR

This paper proposes a decentralized algorithm for assigning electric vehicles (PEVs) to charging stations (CSs) based on individual driver preferences. It leverages a permissioned Distributed Ledger Technology (DLT) to enforce compliance and ensure a reliable charging experience, addressing limitations of previous centralized and less realistic approaches.

What Problem Does This Paper Address?

The increasing adoption of electric vehicles (EVs) presents a challenge for power grids and urban mobility. Efficiently assigning PEVs to charging stations (CSs) is crucial to avoid grid strain and long queues. Existing solutions often rely on centralized systems, assume unrealistic driver cooperation, or fail to account for diverse driver priorities (e.g., cost, time, distance). This paper tackles the problem of optimally assigning PEVs to CSs in a decentralized manner, considering personalized cost functions and introducing a novel DLT-based compliance mechanism to ensure drivers adhere to their assignments.

Methods

The core of the proposed solution is a decentralized stochastic algorithm for PEV-CS assignment. Each PEV broadcasts its location, personalized cost function (a weighted sum of charging time, price, and distance), and energy needs. Charging stations then broadcast a "green signal" with a frequency proportional to how well they match the PEV's cost function. This allows the PEV to select the most convenient CS without direct communication with other PEVs or CSs.

A key innovation is the use of a permissioned DLT, specifically a Directed Acyclic Graph (DAG) like the IOTA Tangle, to enforce driver compliance. This DLT acts as a "compliance layer" where PEVs deposit a token bond when accepting a CS assignment. This bond is returned if the driver reaches the CS within a specified time. A feedback controller regulates the value of these token bonds to maintain a desired compliance level. This approach moves beyond game-theoretic incentives to a more direct economic commitment mechanism.

Datasets

The paper utilizes simulations for evaluation.

  • Mobility Simulator SUMO: Used for simulating vehicular flows in realistic urban networks, specifically for the city center of Pisa, Italy.
  • Monte Carlo Simulations: Used to average results over multiple runs and different configurations, particularly for analyzing the effectiveness of multi-objective cost functions and compliance levels.

No new benchmarks or datasets were introduced; rather, existing simulation tools and methodologies were employed.

Key Insights and Ideas

  • Personalized Charging Preferences Matter: Drivers have diverse priorities (cost, time, distance), and a one-size-fits-all assignment strategy is suboptimal. The algorithm effectively incorporates these individual preferences.
  • Decentralization Enhances Scalability and Privacy: By avoiding direct communication between PEVs and CSs, the system reduces communication overhead and protects sensitive information (e.g., queue lengths, energy needs).
  • DLT for Real-World Compliance: The novel application of DLT as a token-bond mechanism provides a robust and economically grounded way to enforce driver compliance, a critical factor often overlooked in previous work.
  • Compliance is Crucial for Service Quality: Simulations demonstrate that even a small fraction of non-compliant users can significantly degrade the performance of the assignment system, highlighting the importance of the proposed compliance layer.
  • DAGs as a Suitable DLT for IoT: The paper argues that DAG-based DLTs are more appropriate for smart city applications than traditional blockchains due to their lower transaction fees and higher throughput, making them suitable for microtransactions and real-time data exchange.

Key Results

  • Decentralized vs. Centralized Performance: The decentralized stochastic algorithm achieved performance very close to the optimal centralized deterministic solution, with an average charging time only about 5% higher (8 minutes vs. 7.6 minutes) and an RMSE of 0.5087 minutes.
  • Impact of Cost Function Weights: Simulations showed that optimizing solely for charging costs led to significantly longer charging times (around 110 minutes) as all vehicles converged to the cheapest CSs. Conversely, optimizing for charging time alone resulted in much shorter times.
  • Realistic Urban Simulation (Pisa):
    • Without the proposed assignment procedure, vehicles tended to cluster at cheaper CSs, leading to uneven utilization.
  • With the proposed procedure, a more uniform utilization of CSs was achieved, with expensive CSs becoming attractive when queues formed at cheaper ones.
  • Compliance Impact: Average charging time decreased monotonically as the compliance level increased, underscoring the effectiveness of compliance mechanisms in improving service quality.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the decentralized solution is probabilistic and not guaranteed to find the absolute optimal CS, though it is very close.
  • The assumption of a short time for the assignment optimization problem to be computed (1 second) might be challenging in highly dynamic scenarios.
  • The current work focuses on the DLT for compliance; its broader use as a general communication layer is mentioned but not fully explored.
  • Future Work:
    • Practical implementation of the proposed algorithm, including a permissioned DLT and a centralized version for comparison.
  • Development of a smartphone app for PEV drivers to use the system.
  • Involving CSs in advertising their availability using private information (e.g., queue lengths).
  • Further exploration of the DLT system for broader communication tasks within a smart city.

Summary updated

BibTeX
@article{moschella2021decentralized,
  author = {Michela Moschella and Pietro Ferraro and Emanuele Crisostomi and Robert Shorten},
  title = {{Decentralized Assignment of Electric Vehicles at Charging Stations Based on Personalized Cost Functions and Distributed Ledger Technologies}},
  journal = {IEEE Internet of Things Journal},
  year = {2021},
  doi = {10.1109/jiot.2021.3052045},
}
Figure