ABSTRACT
This chapter focuses on the prediction of vehicle trajectory that could be used as part of a vehicle-to-vehicle (V2V)/vehicle-to-infrastructure (V2I) collision avoidance system. The mathematical estimation algorithm selected was the Kalman filter (KF), using multiple prediction models in a multiple models framework. The KF models were created based on the possible spatial states a vehicle could be found in, such as constant location, constant velocity, constant acceleration, and constant jerk. The interacting multiple model (IMM) was expected to outperform multiple models adaptive estimation (MMAE), and in the research, it displayed improvements of 20". The dead-reckoning with dynamic error (DRWDE) method showed that a more frequent recalculation of the prediction improved the prediction errors by 25" over the IMM. The chapter focuses on looking for an alternative to vehicle-mounted sensors to predict a vehicle's trajectory to enable older vehicles to participate in a modern V2V/V2I collision avoidance system.
