Traffic jam dissipation capability of human drivers in comparison to intelligently controlled vehicles

In the past, techniques aimed at the dissipation of traffic jams have been realized with different cruise control approaches. Typically, these have been tested and analyzed in an artificial, closed ring road scenario consisting of a circular single-lane road filled with cars. This scenario exhibits similar traffic inefficiencies as real-world scenarios. In the case of human drivers, typical driving behavior may lead to stop-and-go waves and traffic jams. Previous studies have shown that a single intelligently controlled vehicle is already capable of outbalancing these inefficiencies.While different control strategies have been discussed for the intelligently controlled vehicle, this work also analyzes the traffic jam dissipation capability of actual human drivers. Thirty-five human drivers experienced a virtual version of the ring road scenario in a driving simulator. Here, they were surrounded by simulated drivers who behaved according to a car-following model. When traffic obstructions were induced by the simulated vehicles, the participants were tasked with mitigating the occurring traffic jam.An evaluation of the effectiveness of human drivers in alleviating traffic congestion showed markedly different employed strategies, highlighting differences between successful and maladaptive approaches. These results were then compared to various intelligent control approaches, including an imitation learning model trained to approximate a nonlinear model predictive controller.