Safety at Autonomous Intersections under Connection Loss: Connected and Onboard Local Planning

With the emergence of connected autonomous vehicle (CAV) technologies, autonomous intersection management (AIM) has become a preferred approach to manage traffic flows at intersections. In our earlier work, we identified potential execution discrepancies between CAV local planners and global AIM systems. We have also developed a Vehicle-to-Vehicle (V2V) based local planner to enhance safety under perturbation conditions. Building on these findings, this study further assesses safety in scenarios with connection losses by employing onboard motion predictors (OMPs) to compensate for information gaps caused by intermittent connectivity. Results show that our scheme maintains safety under modest connection loss using simple extrapolation prediction replacement while incorporating advanced OMP helps near-miss scenarios under severe connection loss.