Verifiable and Robust Monitoring and Alerting System for Road Safety by AI based Consensus Development on Blockchain

The integration of AI and the Blockchain has been explored by existing solutions that cooperatively detect and alert a dangerous situation for road safety purposes. However, most such solutions endorse a transaction without considering the context conveyed. Alternatively, their concept has not been implemented as a deployable system for a trial in the field. Considering a use case to be the safety of a pedestrian in a traffic junction, this paper proposes a novel approach, the VErifiable and Robust Monitoring and Alerting (VERMA) system, that 1) detects a potentially dangerous situation (like a fallen pedestrian on the road), 2) immediately alerts the other stakeholders, the driver, and the driving assistance system in the proximity, and 3) records the detected danger in the Blockchain with auditable evidence. The key feature of the proposed system is the multi-AI PBFT-based voting mechanism that enables the other participating nodes to verify the context of the reported dangerous situation using their own AI from different angles and locations in the junction. The VERMA system was implemented by emulating a pseudo vehicle using a smartphone and a GPU-enabled laptop computer that satisfies the requirement to be tested in the field. The evaluation results through the field trial confirm the end-to-end integration of the proposed system with the fast consensus development. This paper also organizes the lessons that should be considered for real-world deployment.