Authority Is Not Necessary: Packet‐Level Evidence for Load‐Dependent Contention‐Window Control in Unsaturated V2I DCF

ABSTRACT Adapting the minimum contention window CW min to the contending‐vehicle population is a standard prescription for IEEE 802.11 DCF in vehicle‐to‐infrastructure (V2I) networks. We first formalize, within a decoupled analytical model, the condition under which the window becomes the binding control variable—an authority criterion—and evaluate the margin ℳ = τ sat /τ load separating the backoff‐limited and load‐limited attempt‐rate branches at the airtime‐capacity boundary. At the reference window CW min = 32 that margin is 12.5–25.1 across contending populations of 30, 40, and 60 vehicles, and closing it analytically requires CW min ≈ 432–968, an order of magnitude above that reference. Taken at face value, this appears to rule out useful contention‐window control below that threshold. Packet‐level ns‐3.42 simulation of the same 802.11p/10 MHz configuration shows otherwise. Holding the vehicle population fixed at N = 40 and sweeping the minimum contention window across offered loads approaching the airtime‐capacity boundary, we find that CW min measurably shapes the collision process, one‐way delay, and age of information long before the analytical branches cross. The tested window that minimizes age of information rises from approximately 127 at λ = 12.8325 packet/s/vehicle to a statistically indistinguishable region of roughly 191–383, whose lowest point estimate is 255, at λ = 13.5454 packet/s/vehicle, although the population is unchanged. A density‐only controller—transplanting the linear CW‐versus‐neighbor‐count rule of Stanica et al., calibrated for unacknowledged VANET beaconing with a contention‐window ceiling of 50, into this acknowledged‐unicast setting with a ceiling of 1023—returns a fixed CW = 20 regardless of load, well short of the tested optimum at either load point and therefore cannot track the shift. The results reconcile the two views: The analytical authority margin remains a valid diagnostic of branch separation within the decoupled mean‐field model. Branch crossing, however, is not a necessary condition for the contention window to exert practically significant control at packet level, and vehicle density alone is not a sufficient state variable for setting it.