Spatial Positioning Token (SPToken) for Smart Mobility

Roman Overko, Rodrigo H. Ordóñez-Hurtado, Sergiy Zhuk, Pietro Ferraro, Andrew Cullen, Robert ShortenIEEE Trans. ITS 2019

Introduction

This paper introduces SPToken, a permissioned distributed ledger technology (DLT) framework designed to facilitate "smart mobility" without compromising user privacy or disrupting natural traffic flow. By utilizing virtual tokens that "hitchhike" across real-world vehicles, the system enables reinforcement learning (RL) to map traffic patterns and identify optimal routes without influencing driver behavior or creating artificial congestion.

Problem Statement

Modern navigation platforms like Waze or Google Maps face two primary challenges:

  1. Data Sovereignty: Users typically surrender ownership of their location data to corporations without built-in mechanisms for retention or compensation.
  2. The "Closed-Loop" Problem: In traffic management, the act of measuring the environment can inadvertently alter it. If an RL algorithm directs "probing" vehicles to specific streets to assess congestion, those vehicles may themselves cause the congestion they intend to measure.

The authors propose a non-intrusive sampling method that is secure, private, and does not perturb the environment it seeks to optimize.

Methodology

The SPToken architecture is built on a multi-layered approach:

  • IOTA-based DLT: Utilizing a Directed Acyclic Graph (DAG) structure allows for high-frequency, fee-less microtransactions suitable for urban IoT environments.
  • Proof-of-Position (PoP): Physical sensors at junctions verify a vehicle's presence, preventing "Sybil attacks" or the injection of fake data.
  • Adaptive Proof-of-Work (aPoW): To ensure data integrity, vehicles submitting data that deviates significantly from the consensus must solve more complex mathematical puzzles, making malicious data entry computationally expensive.
  • Virtual "Jumping" Tokens: This core innovation decouples the RL agent from the physical vehicle. A virtual token "jumps" between vehicles to complete its probing mission. For instance, if a token needs to explore a specific turn that a car is not taking, it can wait for another vehicle to carry it through that segment.
Figure 3: The virtual token 'hitchhikes' on vehicles to explore routes (blue) that may differ from the driver's intended path (orange), allowing for non-intrusive traffic exploration.

Experimental Setup and Datasets

The system was validated using the SUMO (Simulation of Urban Mobility) simulator applied to a digital model of South Dublin, Ireland. To make the RL state space manageable, the researchers performed State Merging, condensing 10,803 individual road links into 3,663 logical states by grouping segments where no turning choices exist.

Figure 4: The state-merging process simplifies the complex urban road network into a manageable Markov Decision Process (MDP) for reinforcement learning.

Key Results and Insights

  • Learning Efficiency: In simulations featuring intermittent traffic jams, SPToken successfully identified and avoided congestion, reverting to the shortest path once the jam cleared.
  • Scalability: The speed of learning follows an exponential-like decay relative to the number of tokens. Increasing the token count significantly reduced the number of episodes required to converge on an optimal policy.
  • Reliability: Using 80 independent tokens resulted in a narrow 95% confidence interval, demonstrating high reliability in converging on the ground truth of the traffic environment.
  • Algorithm Agnostic: The architecture proved flexible, working effectively with both the UBEV algorithm and UCB Q-Learning.

Conclusion and Future Work

SPToken demonstrates that it is possible to decouple traffic exploration from physical action, allowing a system to learn "what-if" scenarios without perturbing traffic density. Future research will focus on quantifying network latency in real-world DLT deployments and moving from simulations to a physical testbed at University College Dublin to evaluate hardware performance in the field.

Summary updated

BibTeX
@article{overko2019spatial,
  author = {Roman Overko and Rodrigo H. Ordóñez-Hurtado and Sergiy Zhuk and Pietro Ferraro and Andrew Cullen and Robert Shorten},
  title = {{Spatial Positioning Token (SPToken) for Smart Mobility}},
  journal = {IEEE Trans. Intell. Transp. Syst.},
  year = {2019},
}
Figure