Modeling and field experiments on autonomous vehicle lane changing with surrounding human-driven vehicles

Autonomous vehicle (AV) technology is widely studied in both industrial and academic communities since it is regarded as a promising means for improving transportation safety and efficiency. Lane changing is a critical link for higher-level AV operations. However, few studies on AV lane changing consider the dynamics of surrounding vehicles, particularly in a mixed traffic environment including human-driven vehicles (HVs). Therefore, this article presents a dynamic lane-changing model for AV incorporating human driver behavior in mixed traffic. The proposed model includes four key components: car following (and lane keeping), lane-changing decision, dynamic trajectory generation, and model predictive control (MPC)-based trajectory tracking. AV longitudinal control algorithm is also depicted in detail in this article. Field experiments are conducted on a large-scale test track to test and validate the proposed model. An AV and three HVs are used in the lane-changing experiments. Different human driver behaviors are considered in the experiment settings. Experimental results show that the proposed lane-changing model can complete lane-changing maneuvers efficiently when HVs are cooperative and can also robustly abort them when HVs are uncooperative. Compared with the measured human lane-changing maneuvers, AV lane-changing maneuvers from the proposed model are more comfortable and safer.