Identification of a linear driver model based on simulator experiments

Abstract

The paper presents a driver model identification method based on simulator experiments. The visual and vestibular perception of the driver can be modeled by measuring proper signals of the vehicle motion and the environment during the real-time driving of the simulator. The parameters of different drivers are then estimated using a linear difference autoregressive model structure and least-squares estimation techniques. The aim of the identification is to describe the different driver behaviors with the parameters of a driver model with similar structure. The identified driver models are validated by simulation using the same excitation signals as in the simulator experiment and comparing the measured and simulated output of the driver. The main novelty of the paper is the identification method in which a real-time simulator is used.

Real-time characterisation of driver steering behaviour

“…Interestingly, they adopt a simpler structure of driver model than most, which is essential to support well-determined parameter identification. Papers with a similar focus include identification of driver delay by Hosseini et al [18] and identification of a simple second order model by Mihaly and Gaspar [19]. Neither of these considers real-time adaptation which would also be needed to adapt to driver changes, though [18] uses least-squares identification which could be made recursive.…”

Real-time Look-ahead Cruise Control Simulator

“…A hardware-in-the-loop (HIL) simulator based on a real vehicle was introduced in (Szalay et al, 2012), where a high-accuracy validated simulation software was attached to several hardware components. This simulator has already been used to set up longitudinal and lateral driver models based on driver experiments, see (Mihály and Gáspár, 2014). Also driver behavior, in terms of longitudinal dynamics, has been investigated and compared to the proposed look-ahead method, see Mihály et al (2012;.…”