Cooperative decentralized intersection collision avoidance using Extended Kalman Filtering

Vehicle accidents are a major cause of death and claim more than 40,000 lives annually in the US alone. A substantial portion of these accidents occur at intersections. Stop signs and traffic signals are some of the intersection control devices used to increase safety and prevent collisions. However, these devices themselves can contribute to collisions, are costly, inefficient, and are prone to failure. This paper proposes an adaptive, decentralized, cooperative collision avoidance (CCA) system that optimizes each vehicle's controls subject to the constraint that no collisions occur. Three major contributions to the field of collision avoidance have resulted from this research. First, a nonlinear 5-state variable vehicle model is expanded from an earlier one developed in. Second, a set of constrained, coupled Extended Kalman Filters (EKF) are used to predict vehicular trajectory in real-time. Third, a vehicular network based on the new WAVE standard is presented that provides cooperative capabilities by enabling intervehicle communication. The system is simulated against today's common intersection control devices and is shown to be superior in minimizing average vehicle delay.