Development of a Dangerous Driving Suppression System Using Inverse Reinforcement Learning and Blockchain

Abstract

Casualty injury rate in car accident is still high level. The number of annual traffic accident casualties in the world today is as much as 1.35 million, and those accidents are caused by reckless driving such as signal ignoring and over speed. In this research, we propose a system which can encourage drivers to make safe driving voluntary using a driving manner evaluation mechanism. Our proposed system uses both inverse reinforcement learning and block chain platform. As for the system development environment, we use a small robot car with a camera attached to the front of the car, and operate on a test course simulating a single lane road. Using the image from the camera, each state corresponding to the image is evaluated and reward value is assigned using inverse reinforcement learning. Either giving reward according to the evaluation value or creating rankings by verifying whether the driving accuracy is improved, the proposed system can make good motivation with competitive spirit.

Preliminary subjective test was performed with 9 subjects who drove a small vehicle. The test result shows positive feedback in case of both giving rewards and giving better ranking. ANOVA result shows that there is a significant difference at a significance level of 5%.