Mimicking human driving behaviour for realistic simulation of traffic flow

Simulated autonomous vehicles running on simplified driver models may lead to some drawbacks because of the very limited 'human-like' errors and driver?driver interactive capability. This study proposes a simulation framework of realistic traffic flow that contains several types of autonomous vehicles. The vehicles operated by ideal driver models simulate the traffic flow in the far-field areas while the vehicles operated by cognitive driver models simulate the near-field areas owing to more 'human-like' behaviour. Extended human-in-the-loop simulation uses real human drivers to enhance active driver-driver interaction. This simulation framework can be used to improve the fidelity of driving simulators.