Turning Motion Planning for Connected and Automated Vehicles at Unsignalized Intersections Considering Multiple Left-Turn Exit Lanes

Abstract

Left-turn maneuvers pose significant challenges for connected and automated vehicles (CAVs) regarding both safety and traffic efficiency, particularly at mixed traffic unsignalized intersections. This paper addresses these challenges by proposing a multiple turning exit lanes (MTEL) approach. Inspired by behaviors of human drivers observed at unsignalized intersections, the MTEL approach aimed to optimize the turning paths of left-turn vehicles by using multiple available exit lanes. Given the uncertain intentions of surrounding human-driven vehicles (HVs), the decision-making processes for left-turn CAVs were formulated as a partially observable Markov decision process (POMDP) and addressed using an adaptive belief tree (ABT) algorithm. The results demonstrate that the MTEL approach provides an effective turning strategy for left-turn CAVs. Compared to conventional fixed-path approaches, the MTEL method achieved significant improvements, including a reduction in travel time by 10.9% and stops by 49.5% across all experimental scenarios. The proposed method proved effective in identifying feasible turning paths amid mixed traffic conditions, thereby enhancing the efficiency of left-turn CAVs while maintaining safety.