Research on monitoring mechanism of autonomous taxi: an evolutionary game approach

As an emerging mode of transportation, autonomous taxis have attracted widespread attention from scholars in various fields. However, theoretical research on how to monitor their operational safety still remains scarce. In light of this, we construct a tripartite game model involving safety regulatory department (SRD), autonomous taxi company (ATC), and passengers using evolutionary game theory (EGT), so as to explore the dynamic evolution of safety supervision mechanism within the entire system. Simulation results indicate that in the basic model (with fixed incentives), the system fails to stabilize at any equilibrium point, implying that a robust regulatory mechanism can never be established. Consequently, we delve into the effects of three dynamical incentive mechanisms: dynamical punishment control, dynamical reward control, and dynamical reward-punishment control. Our findings reveal that dynamical punishment control is detrimental to the stabilization of passenger supervision strategies, while dynamical reward control is highly sensitive to changes in reward parameters. In contrast, the dynamical reward-punishment model demonstrates robustness, contributing to the establishment of a safety supervision system jointly maintained by passengers and self-driving taxi companies. We hope this work provides novel theoretical insights into how to substantially improve the safety of autonomous taxi services.