This paper focuses on acoustic road traffic monitoring and looks, more specifically, into the problem of speed and wheelbase length estimation of two-axle vehicles as they pass by. It is known that both front and rear axle trajectories may be dissociated using cross-correlation based methods in conjunction with a well designed two-element microphone array placed on the roadside. This is mainly due to the broadband nature of the tyre/road noise which makes two peaks appear, one per axle, in the correlation measurement when the vehicle is in the broadside direction. This paper aims at analyzing such a “bimodal’’ observation in order to automatically extract the position, speed, and wheelbase length of passing-by vehicles. We propose to conduct this tracking problem using a particle filter that model the position-variant bimodal sound source nature of the vehicles. The theoretical developments presented in this paper are experimentally assessed through real in-situ measurements.