Secure Outdoor Smart Parking Using Dual Mode Bluetooth Mesh Networks

Efficient parking lot automation continues to be a focal point of smart city initiatives. Most existing unattended parking lots suffer from a lack of seamless automation, instead deploying ticketing and payment at ingress and egress points or other systems with heavy user involvement that often form bottlenecks. Similarly, many lots use per-space sensing with expensive networking and power requirements simply to determine space occupancy. Parking solutions that are free from the delay caused by this user burden and infrastructure could experience faster occupancy turnover with lower cost. An ongoing challenge in developing seamless parking experiences is the detection and identification of vehicles in parking spaces without the need for complex and expensive per-space occupancy detection technology. We develop a smart parking solution that uses a single low-power wireless radio technology to seamlessly perform parked vehicle localization and transport of sensor data for use by a central management system. Our solution uses a sparse, self-forming network of dual-mode Bluetooth sensors within a parking area to observe the presence of customized authenticated Bluetooth Low-Energy (BLE) beacons placed in vehicles parked in the lot. Our localization technique is based on radio fingerprinting using Received Signal Strength Indication (RSSI) values from the beacon, and a random forest machine learning classifier that predicts where the vehicle is parked based on its fingerprint. We implemented our solution in Python on commodity Internet of Things (IoT) hardware and deployed it to a 105 space outdoor parking lot. There, we conducted fingerprinting and prediction experiments. Our results show that our exact-space prediction model evaluates with a high accuracy using radio training data (90.7\% correctly identified), and our in-vehicle tests show a promising result (69.17\% accurate up to and including 3 spaces away), even without employing tuning and data filtering techniques. This encouraging result shows that localization using Bluetooth is a viable means of managing parked vehicles, with great promise for a variety of future parking management applications.