Integration of Intelligent Driver Model with Interaction-Aware LMB (IA-LMB) Filter for Vehicle tracking.

Vehicle tracking is one of the most common applications of multi-object tracking. In the context of Bayesian filtering, including accurate information about the tracking scenario tends to improve tracking performance. However, in some cases, including more information increases the complexity of the tracking system. This paper presents a novel method of direct mathematical incorporation of accurate motion models for vehicles, specifically the intelligent driver model (IDM), within the interaction-aware labeled multi-Bernoulli (IA-LMB) filter. The proposed method provides the inclusion of a more accurate vehicle motion model into the IA-LMB filter without increasing the complexity of the IA-LMB filter. The proposed filter outperforms traditional methods, which do not consider interactions between targets such as the PHD filter and the LMB filter. Moreover, the proposed method outperforms the original IA-LMB filter for the interacting vehicles owing to more accurate modeling of vehicle motion, and hence the inclusion of interactions. The performance of the proposed method has been evaluated using the optimal sub-pattern assignment (OSPA) metric for the overall simulation and the mean square error (MSE) for the interacting vehicles.