Abstract
<jats:p>The dependence of capacity gains and hysteresis suppression on the choice of autonomous-vehicle (AV) behavior rule is investigated using a Nagel--Schreckenberg-type cellular-automaton model of mixed AV/human-vehicle (HV) traffic. Four AV following rules---naive (deterministic Nagel--Schreckenberg), IDM, ACC, and CACC---are implemented under a common collision-safety architecture, in which each vehicle computes a desired speed from its own rule and is subject to the same hard safety floor. Slow-to-start (VDR) dynamics are retained for HVs to induce hysteresis, and two independent detection methods are used to quantify the hysteresis loop width as a function of AV penetration rate $r$. The framework is validated against an exact two-species disordered-TASEP solution at $v_{¥max}=1$, showing agreement to within 0.53¥%. The capacity gain from AV penetration is found to be highly sensitive to the calibration of the AV following rule rather than being an intrinsic property of the rule itself: under human-calibrated IDM parameters, AV penetration lowers capacity below the human-only baseline at every penetration rate tested, whereas doubling the acceleration/deceleration capability reverses this shortfall into a capacity gain. CACC, in contrast, saturates the common safety floor and becomes indistinguishable from the naive rule at high $r$, while ACC remains distinct owing to its longer effective headway. Furthermore, hysteresis suppression is shown to be threshold-like rather than proportional to $r$: the loop width remains largely unchanged up to $r¥approx0.4$ and vanishes only as $r¥to1$, indicating that AVs must dominate the population of stopped vehicles before slow-to-start asymmetry is eliminated. These results demonstrate that the widely assumed benefit of AV penetration for traffic flow is not a model-independent conclusion, but instead depends sensitively on the specific behavior rule and its calibration.</jats:p>