Mixed-autonomy traffic instability peaks at 45 percent CAV penetration as cut-ins exploit defensive AV behavior

The study integrates a game-theoretic friction term into the macroscopic kinematic wave model to bridge microscopic interactions and network-level flow. Authors derive a closed-form parameter that modifies the conservation law, showing how exploitable merges generate perturbation source terms that reduce capacity even without stochastic driver error. Numerical simulations confirm the convex efficiency curve with a clear instability regime at mid-penetration levels.

Real-world data from mixed fleets on highways already show elevated lane-change conflicts when automation penetration sits between 30 and 60 percent, yet regulators still treat penetration as a linear safety benefit. The model highlights that early CAVs programmed for risk aversion amplify rather than dampen these conflicts, a dynamic absent from most current NHTSA and EU safety assessments.

Without policy that penalizes strategic exploitation or mandates cooperative CAV control, intermediate deployment phases may degrade throughput and increase rear-end risk. Follow-on empirical validation requires instrumented corridors that log both merge timing and AV controller responses at scale.

Next steps include embedding socially aware reward functions in AV planners and testing whether mandatory minimum headways or merge-priority protocols flatten the instability peak.