Real-Time Position Tracking and Velocity Estimation of Moving Vehicles Using Lidar Data

Niamke Giraud (University of Massachusetts, Dartmouth)
George Gunter (University of Illinois at Urbana-Champaign)
Yuriy Slashchev (Stony Brook University)

In the work performed by Stern et al. (2017), it was shown that a low penetration of autonomous vehicles (AVs) into traffic flow is able to dampen phantom traffic waves by means of the AV maintaining a control velocity dictated by distance from the lead vehicle. Detection of the lead vehicle was performed only through finding the closest point in 2D LIDAR data from the AV. This form of detection may not provide an accurate estimate of the location of other vehicles in the road environment, which is required for maintaining a correct control velocity to dampen the traffic flow shock waves. In this work, we implement an algorithm that performs vehicle recognition and position estimation on 3D LIDAR data, followed by tracking and relative velocity estimation on the acquired position data. Vehicle recognition is performed on the 3D data by means of blob detection and classification into vehicle and non-vehicle objects. A linear Kalman Filter is used with the centroid information extracted from the 3D data in order to account for uncertainties in position measurements and to perform state estimation of other vehicles by means of a constant acceleration dynamical model. The Hungarian algorithm for assignment performs matching between vehicle detections and previous state estimations. This method allows us to detect and track other vehicles with increased precision and create a better model of traffic flow.