An Agent-based Reinforcement Learning Model for Simulating Driver Heterogeneous Behavior during Safety Critical Events in Traffic

摘要

Driving behavior in traffic has been modeled quite successfully in simulation software usingpredefined car-following models rules. However, because most car-following models assumethat vehicles could keep a safety distance away to avoid crash related conflicts; they are notcapable to capture naturalistic driving behavior during safety-critical events. Also, vehicledetailed lateral maneuvering have not been simulated in most simulation software. The proposedmethodology in this paper focuses on establishing a traffic state-action mapping rule to simulatereal driver actions including risky behavior that a driver would take during safety critical eventsinstead of the predefined actions by car-following models. To analyze individual drivercharacteristics and extract driving behavior rules, a fuzzy rule based neural network isconstructed with the objective of presenting driver action rules under associated traffic states. Aspecial training approach Neuro-Fuzzy Actor Critic Reinforcement Learning (NFACRL) isproposed as a methodology to train an agent driver simulator. Vehicle longitudinal and lateralactions are estimated and used as output of this model. The simulated vehicle actions arecompared with naturalistic data.