Rule-Based Decision-Making System for Autonomous Vehicles at Intersections with Mixed Traffic Environment

Before autonomous vehicles are widely spread, they will share the roads with non-autonomous cars. Furthermore, to ensure functional safety of the self-driving cars, the cause of decision-making and control algorithms' failure must be precisely identified. Rule-based methods are promising solutions due to their transparency and comprehensibility. In this paper, a rule-based decision-making system for autonomous vehicles solving a challenge of complex intersection with mixed driving environment is proposed. The system is designed to prioritize road safety and avoid collision with other road users at any cost. The proposed algorithm relies only on available on-board perception and localization sensors, allowing the automated car to operate among human-driven vehicle and without vehicle-to-vehicle communication technology. The system is validated in a simulation study on cross-intersection, where the ego vehicle deals with multiple cars arriving from different sides of the road. The results demonstrate algorithm's robustness and effectiveness under multiple scenarios, when neither intention nor trajectory of other traffic participants is known. Thus, the proposed solution is also potentially applicable to other types of intersections with different traffic rules.