Modeling and capacity analysis of mixed CAV-HV traffic incorporating variable safety distance in a cellular automaton model

With the rapid development of connected and autonomous vehicle (CAV) technologies, highway traffic will remain in a mixed state with human-driven vehicles (HVs) and CAVs coexisting for a long period. Existing studies often assume a fixed random deceleration probability for HVs and apply uniform lane-changing rules, overlooking speed-dependent driving behavior and HV–CAV differences. To address this gap, this study improves the Gipps safety distance rule by introducing a speed-dependent random deceleration mechanism for HVs, and integrates the ACC/CACC models from the PATH laboratory to construct unified car-following and differentiated lane-changing models for HVs and CAVs. The HV model parameters are calibrated using high-resolution NGSIM trajectory data with a genetic algorithm, and a bidirectional four-lane highway cellular automaton environment is developed for simulation. The impacts of CAV penetration rate, reaction time, and lane-changing probability on traffic flow characteristics are systematically analyzed. Results show that when the CAV penetration rate exceeds 40%, road capacity increases significantly and congestion is substantially alleviated. Under high penetration, reducing CAV reaction time from 1.0 s to 0.5 s raises the maximum traffic volume by about 40%. In contrast, higher lane-changing probabilities have limited benefits for capacity but clearly intensify congestion and reduce traffic flow stability. This study proposes a more realistic framework for mixed HV–CAV traffic flow and reveals the mechanisms through which key parameters influence efficiency and stability, offering quantitative insights for CAV strategy design and highway traffic management.