Fig. 7: Driver model.
From: Human-like driving behaviour emerges from a risk-based driver model
A simple driver model that utilises the estimated risk metric to generate control actions is shown. a Driver model control structure: The driver model uses the cost map of the driving scene (information about the environment), and the vehicle states (position: xcar, ycar; heading: ϕcar; and speed: (v) at kth time step to generate the steering angle (δ) and speed (v) for k + 1th time step. b The zoomed-in driver model block: The DRF is a dynamic field and changes its shape with vehicle state, which are inputs to the driver model block. The DRF is multiplied with the cost map of the driving scene and summed over all grid points to generate the quantified perceived risk (cost function). The driver model algorithm uses the computed cost function, and the vehicle states to generate the speed (v) and steering angle (δ) for next time step. The DRF model algorithm is based on the risk-threshold theory and compares quantified perceived risk (C) with risk threshold (Ct). The DRF can be individualised based on DRF parameters while the driver model parameters determine how the cost (perceived risk) is converted to control actions (speed and steering). c Driver model algorithm: At each time step (k), we compare the risk (Ck) to risk threshold (Ct), and speed (vk) to the goal (Vdes). This results in four distinct cases of inequality.
