Analysis and comparison of driver intent identification methods

This thesis describes and analyses methods for the driver behaviours prediction in an urban intersection. The principal prediction methods found in literature are considered: Bayesian Approach, Hidden Markov Model and Interacting Multiple Model. For each method the features, equations and algorithms are discussed. The algorithms are tested on simulated and real trajectories, comparing the methods' results in terms of prediction horizon for a set intent probability. On the real trajectories a sensitivity and specificity analysis is also performed.