Recent transportation research often suggests that autonomous vehicles (AVs) can improve traffic flow efficiency, as they are able to eliminate misconduct and confused mindset of human drivers. Nevertheless, as a special class of ground robots, autonomous vehicles are inevitably subject to robotic errors in their operations, particularly in the perception module. This causes inconvenient uncertainties in their movements, or even car collisions. Consequently, conservative operational strategies, such as larger headway and slower speeds, are employed when operating automated functions on roads, sacrificing traffic capacity for safety. To reconcile the inconsistency, this paper proposes an analytical model framework that delineates the endogenous reciprocity between traffic safety and effi...[ Read more ]

Recent transportation research often suggests that autonomous vehicles (AVs) can improve traffic flow efficiency, as they are able to eliminate misconduct and confused mindset of human drivers. Nevertheless, as a special class of ground robots, autonomous vehicles are inevitably subject to robotic errors in their operations, particularly in the perception module. This causes inconvenient uncertainties in their movements, or even car collisions. Consequently, conservative operational strategies, such as larger headway and slower speeds, are employed when operating automated functions on roads, sacrificing traffic capacity for safety. To reconcile the inconsistency, this paper proposes an analytical model framework that delineates the endogenous reciprocity between traffic safety and efficiency that arises from robotic uncertainty in AVs. Car-following scenarios are extensively examined, with uncertain headway as the key parameter for bridging the capacity of a single lane and collision probability. A two-state stochastic process derived from the dynamics of the discrete AV system is then introduced to describe the passability of a lane, where the collision-inclusive capacity is obtained to serve as the ultimate performance measure of fully autonomous traffic. With the help of this analytical model, it is possible to support the setting of critical parameters in AV operations and incorporate optimization techniques to improve related managerial strategies and policies. On the other hand, under the quantitative connection of this method, safety and efficiency objectives for the transportation system provide more precise targets and directions for the design and development of autonomous vehicles. Moreover, the generalization of this model is presented, providing the possibility to explore more complex AV models and scenarios.