Merge-Assist Feature for Connected and Automated Vehicles in Mixed Traffic, 03-R6525

Background

Highway merges and interchanges are challenging maneuvers for human drivers as well as for automated (autonomous) systems. The reasons are that they are (a) safety critical and (b) they require negotiation between independent agents impacted by many factors, such as road configuration, vehicle sizes, visibility/occlusions, etc. For this reason, literature and software packages on automated merges mostly rely on established priority, e.g., first-in-first-out (FIFO) priority. The drawbacks include a need for a roadside coordinator, long range for vehicle-to-vehicle communication (e.g., 300m or more), and potential string instability where each subsequent vehicle brakes a bit later and a bit more than the preceding one.

Approach

In this work we considered a decentralized, control barrier function-based approach that avoids prioritization. The instability-driven algorithm (IDA) combines safety filter functionality and negotiation capability into one providing agility/responsiveness to a group of merging vehicles. This is akin to a fighter jet aerodynamic instability improving maneuverability. The approach is based on an algorithm proposed for robots (by the PI), but its application to merge coordination is original.

Accomplishments

Analysis and extensive Monet Carlo simulations have shown that, compared to alternatives, the IDA is more agile, faster, and more robust to unexpected behavior, reduced vehicle communication range (works well at 80m range), and latency. In the one degree-of-freedom case (merging 2 lanes into 1), the method produced 23% fuel economy improvement over FIFO. In the two degrees-of-freedom case, such as the interchange shown in Figure 1, IDA vehicles each maintained safety through barrier constraints and always completed the lane swaps in time even though they did not know if other vehicles were changing lanes or not.