Autonomous overtaking in highways: A receding horizon trajectory generator with embedded safety feature

Abstract

A primary task in autonomous driving is to design a control algorithm that presents an effective and yet human-compatible behavior. To this aim, present paper considers the problem of autonomous overtaking under both state and environment constraints with dynamic surrounding vehicles. The solution is formulated based on quadratic programming optimization and is solved in receding horizon fashion. The proposed method evaluates the traffic condition online and executes a safe trajectory of an overtaking maneuver with on-coming traffic. To better incorporate traffic participants’ behaviors, a dynamic predictive model-based reachability analysis of the surrounding vehicles is utilized in the design. Reachable sets aim to ensure the drivability of the planned motions, as well as producing drivable collision-free trajectories. To this end, forward reachable sets are employed by predicting traffic vehicles’ future actual and worst-case behaviors in the design, in which the autonomous vehicle determines its trajectory accordingly. A simulation scenario is tested via MATLAB/Simulink for autonomous overtaking and the effectiveness of the proposed method is shown, demonstrating the potential utility of the present approach for implementation as an advanced driver assistance system (ADAS) in next-generation vehicles.

Keywords

• Autonomous overtaking
• Quadratic programming
• Model predictive control
• Reachable sets