Competition Threaten Risk Degree Assessment Strategy in Lane-changing Driving Scenarios based on Non-cooperative Game Theory for Autonomous Systems

A system for lane-changing driving scenario analysis and the CTRD (Competition Threaten Risk Degree) assessment based on Non-cooperative Game Theory is proposed. By computing with the real-time collected direct and indirect information, shallow and deep strategy analysis is implemented with uncertainty consideration, and the CTRD of each sampling of competing actions is calculated. In the ongoing potentially critical scenario with high occurrence rate, it could support the decision of whether autonomous driving strategy system is competitive or compromised, and the enhancement of the driving safety to avoid clash accident by rational compromise, or could keep the traffic efficiency with appropriate competition operations after beneficial consideration and risky consequences affordability calculation. The experiment with more than 3000 groups real-road cut-in data shows the relevance of each interactively tentative and feedback action in continuous non-cooperative competition rounds. The SVM method is employed to compare the result of driver’s vehicle operation after facing the theaten challange between CTRL-based strategy prediction and real reactions, the coincidence rate is 94%.