Carefull Tread: A Lightweight Consensus in Blockchain to Trust Mobility Data using Q-Learning

Mobility data generated from Connected and Autonomous Vehicles (CAVs) and micro-mobility devices (e-scooters, e-bikes and smartphones) is critical to understand different safety and future Intelligent Transportation System (ITS) needs. Blockchain is used to endorse such mobility data in a secure manner, however, trusting large amount of data from massive number of devices is challenging as results in scalability issues along incurring longer delays. To cater this, this paper employs reinforcement learning, specifically, Q-learning towards a lightweight mobility data validation at scale. We derive the optimal policy for a node to endorse mobility data generated by nearby nodes in a blockchain network. We present a novel consensus protocol for the node to learn the block sampling rates and time delay thresholds based on the neighborhood topology and network connectivity. We simulate the proposed consensus protocol in NS3, and it has shown to achieve lower delays, overhead and high throughput for up to 50 nodes simultaneously endorsing each others mobility data in New York city.