Optic flow and geometric field of view in a driving simulator display

Abstract

Conditions of high and low optic flow were generated in a display used for a driving simulator. Thirty drivers were requested to produce vehicle speeds of either 30 or 60 mph with the geometric field of view at 25, 55, or 85 visual degrees. Drivers overestimated the production of 30 by 20 mph. It was also found that the production of speed was highly dependent on the geometric field of view. These results suggest that optic flow presented in a driving simulator display does not correspond with optic flow found in real-world driving.

Introduction

Displays used in virtual reality based driving simulators have special requirements. When a driver/vehicle traverses a modeled 3D environment, the generated optic flow should be similar to that found in the real world [8]. It has been proposed that optic flow can be utilized in a number of driving tasks: (1) to obtain heading information when driving on straight and curved roads, (2) to steer around parked vehicles, roadside trees, etc., and (3) in braking and/or steering to avoid collisions when approaching other moving vehicles [2]. Several researchers have studied optic flow while driving in combination with where a driver looks. It has been found that drivers focus near the point where the roadway meets the horizon [7]. This location is known as the focus of expansion since it appears stationary and optic flow expands from it towards the vehicle. Several researchers have reported that drivers focus on a tangent point when negotiating curves [4], [5], [9]. This tangent point may also be considered a focus of expansion since it is (instantaneously) without motion and optic flow expands from it. The tangent point then moves as a driver traverses the curve. Using a focus of expansion as a “home base” for visual scanning may be optimal in terms of a driver monitoring changes in the pattern of optic flow. When the fovea of the eye is directed at the focus of expansion, optic flow can be detected in the peripheral part of the eye’s retina. The periphery of the retinal region, consisting predominantly of rods, is primarily a detector of motion [3].

In the present study we investigate drivers’ use of optic flow when producing a specified vehicle velocity. In real-world driving, drivers may refer to the vehicle’s speedometer to assist them in obtaining or maintaining a velocity specified on a road sign. Glancing at a vehicle’s speedometer entails suspending monitoring of the forward scene for about three-quarters of a second [6]. Unskilled and impaired drivers could choose inappropriate times to make glances to the speedometer that result in unsafe driving.

Optic flow in a driving simulator is inherently different from that seen in the real world. The huge horizontal width of optic flow in the real world is difficult to duplicate with computer monitors or large screens used by driving simulators. By necessity, many simulators compress the horizontal view in the real world by specifying a virtual scene field of view (FOV) that is greater than the projected FOV. This FOV, which is specified in computer software, is known as the geometric field of view (GFOV). Furthermore, the great variety of displays used in driving simulators, ranging from small computer monitors to large screen displays that provide a 270-deg FOV, make it difficult to compare optic flow among simulator configurations.

It has been suggested that synthetic driving environments should be constructed so that they correspond to real-world experiences, especially with regard to the perception of velocity [1]. Driving simulators, in particular, should accurately portray optic flows found in the real world. This would enable simulator users to accurately estimate and generate vehicle velocities. In this study, we investigate the production of vehicle velocity in a driving simulator as a function of increasing GFOVs and the amount of optic flow.