Experimental Validation of DeeP-LCC for Dissipating Stop-and-Go Waves in Mixed Traffic

We present results on the experimental validation of leading cruise control (LCC) for connected and autonomous vehicles (CAVs). In a mixed traffic situation that is dominated by human-driven vehicles, LCC strategies are promising to smooth undesirable stop-and-go waves. Our experiments are carried out on a mini-scale traffic platform. We first reproduce stop-and-go traffic waves in a miniature scale, and then show that these traffic instabilities can be dissipated by one or a few CAVs that utilize Data-EnablEd Predicted Leading Cruise Control (DeeP-LCC). Rather than identifying a parametric traffic model, DeeP-LCC relies on a data-driven non-parametric behavior representation for traffic prediction and CAV control. DeeP-LCC also incorporates input and output constraints to achieve collision-free guarantees for CAVs. We experimentally demonstrate that DeeP-LCC is able to dissipate traffic waves caused by car-following behavior and significantly improve both driving safety and travel efficiency. CAVs utilizing DeeP-LCC may bring additional societal benefits by mitigating stop-and-go waves in practical traffic.