Spatial Positioning Token (SPToken) for Smart Parking

TL;DR

This paper introduces SPToken, a novel system combining Distributed Ledger Technology and Reinforcement Learning to guide drivers to parking spaces efficiently using crowdsourced data and virtual tokens.

Contribution

The paper presents a new architecture, SPToken, that leverages DLT and RL with virtual tokens for real-time, distributed parking space estimation without environmental disturbance.

Findings

Simulation demonstrates improved parking search efficiency.

System effectively adapts to changing environments.

Crowdsourced data enhances route estimation accuracy.

Abstract

In this paper, we describe an approach to guide drivers searching for a parking space (PS). The proposed system suggests a sequence of routes that drivers should traverse in order to maximise the expected likelihood of finding a PS and minimise the travel distance. This system is built on our recent architecture SPToken, which combines both Distributed Ledger Technology (DLT) and Reinforcement Learning (RL) to realise a system for the estimation of an unknown distribution without disturbing the environment. For this, we use a number of virtual tokens that are passed from vehicle to vehicle to enable a massively parallelised RL system that estimates the best route for a given origin-destination (OD) pair, using crowdsourced information from participant vehicles. Additionally, a moving window with reward memory mechanism is included to better cope with non-stationary environments. Simulation results are given to illustrate the efficacy of our system.