Optimal coordinated platoon lane change in highways with mixed traffic

Abstract

In the field of connected autonomous vehicles, platoons - where vehicles closely follow one another - have shown promising results in enhancing safety, traffic flow, and fuel efficiency. This study addresses the unique challenges of platoon lane changes, where multiple platoon vehicles have lane change intention and must remain together after the maneuver is completed. We focus on highway environments because many studies have highlighted the advantages of platooning in these settings. We leverage an offline controller synthesis approach to deal with the combinatorial problem of choosing a strategy. Building on concepts from symbolic optimal control, we represent the problem using a weighted directed acyclic graph where nodes are quantized state vectors, edge weights are costs to transition between nodes, and the shortest path solutions represent the optimal platoon lane change strategies. We use a Cached Branch-and-Bound Depth-First Search algorithm to solve the offline control problem due to its anytime capability and low memory requirements. This approach provides real-time decision making and guarantees maneuver success while minimizing completion time or control effort. Previous works either required control of all vehicles on the road, making them inadequate for mixed-traffic scenarios, or fixed the order in which platoon vehicles change lanes, disregarding the current state of surrounding vehicles and maneuver costs. Our framework can describe all previously proposed methods, relies only on cooperation between platoon vehicles, allows for optimization, and produces solutions whose costs decrease as the allowed computational time increases. We employ the VISSIM traffic simulator to compare our approach to the state of the art. The experiments show that we obtain an increase of 13% in the maneuver completion rate along with a decrease of around 15% to 20% in the maneuver completion time and distance traveled to complete the lane change at the cost of an average increase of 2% to 17% in the longitudinal control effort. This trade-off is a direct consequence of having the platoon occupy suitable lane-changing spaces as soon as possible, and this myopic behavior is necessary when there is no information about the future movements of human-driven vehicles. Moreover, the experiments indicate that our approach yields a sharp decrease in cost in relatively short computational times. These results emphasize the potential for deployment of the proposed method in mixed-traffic highway scenarios.