Extended Horizon Tactical Lane Change Planning In Competing Autonomous Vehicles

The main challenge in autonomous driving is to adhere to traffic rules while ensuring safety and security. Lane Change Decisions - deciding how and when to change a traffic lane - are critical in determining traffic safety, travel efficiency and driving quality. Most works in existing literature focus on narrow horizon lane changes, include at most 1 or 2 lane changes, primary reasons being lack of sufficient information, uncertainty, and complexity in planning due to enormous size of possible trajectories of the surrounding vehicles. In this paper, we present a generalized extended horizon (multiple lane changes) tactical lane changing algorithm which 1) can take into account any mobility model for surrounding vehicles and more importantly, 2) also considers the scenario where the surrounding vehicles are also undertaking tactical lane changing in a "smart" way. We show that using a graph theoretic approach to elegantly capture all possible trajectories simplifies and reduces the complexity of the planning algorithm. We present extensive simulation results showing the effectiveness of the approach.