Traffic Estimation and Control in an Era of Mixed Human Piloted and Automated Vehicles

This talk will explore some new directions in estimation and control when the traffic stream is composed of a mix of human piloted and autonomous vehicles. The first part of the talk investigates the problem of modeling and estimating traffic streams in this mixed setting. A connection between the generalized Aw Rascle Zhang model and two-class traffic motivates the choice to model the flow with a “second order” traffic model. With the system dynamics defined, traffic state is estimated via a fully nonlinear particle filtering approach, and results are compared to estimates obtained from a particle filter applied to a scalar conservation law. Numerical experiments indicate that when the penetration rate of automated vehicles in the traffic stream is highly variable, the second order model based estimator offers improved accuracy compared to a scalar traffic flow model. The second part talk explores the problem of controlling the human piloted traffic with only a small number of autonomous vehicles. We modify the experimental setting of Sugiyama et al. (2008) to measure the influence of a carefully controlled AV on human piloted vehicles. Even when the penetration rate of automated vehicles is as low as 5%, we show it is possible to reduce the presence of stop-and-go waves that can appear without the presence of a bottleneck. Our experiments imply that significant improvements in traffic fuel efficiency and safety may be achieved by means of very few mobile actuators in the traffic stream.

Speaker: Daniel Work, Univ. of Illinois, Urbana-Champaign