Predictive Traffic Assignment: A New Method and System for Optimal Balancing of Road Traffic

One of the most prevalent problems for people living in big cities is traffic congestion. To avoid traffic jam, drivers tend to use intelligent traffic-aware route planning service to help them save travel time. While the existing traffic-aware systems independently compute the fastest route based on the current and/or historical traffic condition, they ignore the fact that the uncoordinated decision based on the identical traffic view could lead to new congestion in the future. We propose a novel online route planning system called Predictive Traffic Assignment or PTA that exploits previous planned routes to accurately predict their impact on future traffic. Based on the dynamic prediction, PTA system computes the optimal route for each vehicle. Because it accounts for the impact of the previous vehicles, PTA coordinately assigns different routes to vehicles such that the traffic load is balanced among these routes. We present the models of PTA and implement PTA with an efficient algorithm. We conducted extensive simulation studies based on actual city maps. Simulations show that PTA significantly outperforms the state-of-the-art online and offline approaches. We also show that PTA could result in great saving in travel time, fuel consumption and GHG emissions even if only a small portion of vehicles use PTA.