Automated vehicle control systems need to solve social dilemmas to be disseminated

Abstract

Abstract A new cellular automata traffic model based on the revised S-NFS model was established to consider a mixed flow of automated and human-driven vehicles assuming a multi-lane system. The model further classified automated vehicles into two categories: (1) vehicles with adaptive cruise control and (2) those with cooperative adaptive cruise control that supports so-called platoon driving. A vehicle that favors maximizing individual payoff, which ensures minimizing its own travel time, while maximizing global traffic flux was expected as the entire society. Intensive simulations, wherein automated and human-driven vehicles were presumed as cooperative (C) and defective (D) strategies, respectively, revealed that a D-strategy is always better than a C-strategy to maximize individual payoff as long as a smaller cooperative fraction is imposed. Meanwhile, social optimal could be realized by a situation comprising only automated vehicles. Such a stag-hunt social dilemma implied that an automated vehicle control system (AVCS) cannot permeate into a population of human-driven vehicles if the dissemination stage starts from a single vehicle with an AVCS.