Investigating the effect of controller locations on centralised dynamic traffic control with equilibrium constraints

Abstract

Traffic control policies aim at reducing the negative externalities that ever-growing demand is causing on transportation networks, such as congestion and pollutant emissions. To achieve these goals, policies coordinating and aligning the effects of several individual traffic controllers have received increasing attention in research and development in the past decades. Notwithstanding the considerable efforts, peak-hour congestion is still a global problematic, with commuters spending up to 100 hours per year in traffic jams in major cities like Los Angeles (INRIX’s Global Traffic Scorecard, 2017). In this work we try to assess whether the gap between desired and experienced performance in advanced control policies might indeed be tied to inefficient network design, rather than algorithmic prowess. Based upon our earlier work, we investigate whether a trend can be found between determining locations of controllers in a network following control theoretical insights, and try to confirm our earlier intuitions when dealing with dynamic traffic management, featuring accurate propagation and spillback dynamics. Results comparing the performance of optimal control based on 50 randomly chosen locations and a deterministically chosen set of controllers are shown, highlighting how indeed a strong link exists between controller locations and reachable performance in optimisation based control strategies.