MPC-Based Connected Cruise Control with Multiple Human Predecessors

Abstract

Model predictive control is applied to regulate the longitudinal motion of a connected automated vehicle in mixed traffic scenarios. A prediction method is proposed to enable model predictive control in low-automation, medium-connectivity situations using instantaneous motion information from multiple predecessor vehicles. This includes detection of unconnected vehicles that may be mixed between connected ones. Simulations using real human driver data for the predecessors show that, if the drivers are well-characterized on average, a hidden unconnected vehicle can be detected over 90 % of the time. Moreover, the resulting preview can recover 46 % of the gap in energy performance between single-predecessor prediction and ideal preview. Results are also compared to a classical controller that utilizes instantaneous information from multiple predecessors.

Additional details