Traffic Dynamics at Intersections Subject to Random Misperception

Traffic accidents cause harm to the society. Future technology in autonomous vehicles is expected to eliminate the human factor as one important cause of failure. However, in the near future, autonomous vehicles and human drivers will coexist and downside risk still needs to be tolerated in exchange for mobility. Unsignalized intersections are particularly prone to accidents, as lots of potential conflicts between traffic participants occur. Motorists need to anticipate these on the basis of their perception of the environment and react accordingly. Yet, perceptional errors affect human drivers, and it is important to understand their impact on traffic safety and traffic efficiency. We develop a microscopic model of traffic dynamics at single-lane unsignalized intersections subject to random misperception that may cause accidents. Perceptional errors can be modeled by stochastic processes, e.g., Ornstein-Uhlenbeck processes. We present suitable simulation techniques and characterize the behavior of the traffic system in various case studies. We discuss the impact of errors and safety margins on traffic flow, the number of accidents, and the number of collided vehicles. In terms of perception errors, we consider both homogeneous and heterogeneous traffic participants, reflecting the coexistence of human drivers and autonomous vehicles. The model captures the real-world tradeoff between safety and efficiency for potential future traffic systems.