Abstract
Purpose :
Hemianopic field loss (HFL) may impair detection of hazards, which is a safety concern given that individuals with HFL are permitted to drive in many states. To address this problem, we developed a prototype hazard warning device and evaluated the effects of the device on pedestrian detection using a pedestrian simulator linked to a driving simulator. This novel approach enabled the measurement of responses to (1) human-controlled pedestrians (HCpeds), increasing the relevance to real-world driving where human-human interactions are important; and (2) pre-programmed pedestrians (bots), as used in prior studies.
Methods :
Eight subjects with HFL completed 2 drives with and 2 drives without the device on city routes while gaze was tracked. The device provided a lateralized vibration cue (500ms) from motors embedded in the left or right side of a custom-made seat cover to provide warning of a pedestrian hazard on either the left or right sidewalk, respectively. Responses were analyzed for 16 HCpeds, operated by a trained assistant from the pedestrian’s viewpoint, and 16 pre-programmed bots. HCpeds indicated intent to cross the road by walking to the curb, and then stopping before making a head turn toward the driver; they only crossed if the human operator felt safe to do so. Bots indicated intent to cross by running toward the road; however, they did not stop and kept running irrespective of whether it was safe. The city was populated with other distractor pedestrians in groups and alone. Unsafe responses included collisions, unsafe (harsh) braking and situations where the HCped did not feel safe to cross.
Results :
With hazard warnings, the rate of unsafe responses of HFL subjects decreased significantly (p < 0.001) for pedestrians approaching from the blind (36% to 3%) and seeing (13% to 5%) sides. By comparison, for control subjects, unsafe responses decreased from 7% to 1%. In the phase where pedestrians indicated intent to cross: 1) Hazard warnings increased the proportion of time HFL subjects spent fixating pedestrians approaching from the blind but not the seeing side (p = 0.05); and 2) A significantly greater proportion of time was spent fixating HCpeds than bots (p= 0.01). Finally, all HFL subjects rated the device as helpful.
Conclusions :
Our results provide preliminary evidence that lateralized tactile warning cues can improve responses of drivers with HFL to pedestrian hazards, especially on their blind side.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.