This study evaluates a novel collision warning device to help people with severe vision impairment or blindness avoid collisions with obstacles. The main hypothesis to be tested is that the device reduces the number of collisions with obstacles in everyday activities.

Participants with blindness or severe visual field loss (hemianopia or tunnel vision) will be trained to use a collision warning device that alerts them to impending collisions. Participants will use the device at home during everyday mobility for about 1 month. The device is designed to supplement existing mobility devices, such as a long cane or dog guide. The device will provide warnings about potential collisions with mid- or high-level obstacles (which are typically not detected by a long cane).

Inclusion Criteria:

• Blindness (or very limited vision with visual acuity no better than "counting fingers"),
• Or severe peripheral field loss / tunnel vision (≤ 40° remaining visual field) with visual acuity of at least 20/200,
• Or homonymous hemianopia with visual acuity of at least 20/200;
• Able to walk independently either with or without mobility aids such as a long cane or guide dog (but without the aid of a sighted guide);
• Reports at least minor bumps or collisions within the last 3 months;

Exclusion Criteria:

• Currently participating in a mobility training program
• Diagnosed dementia
• Significant cognitive decline
• Participants with hemianopia will be excluded if they have hemianopia for less than 3 months and/or have spatial neglect

The baseline analysis population includes only those subjects who completed the one-month period of device use and had data successfully recorded by the device.

Adverse event data were collected during the one-month period when subjects were using the device.

A brief questionnaire was administered to each subject by telephone each week during the one-month period of device use to check on how much they were using the device and whether they had encountered any difficulties. One question specifically asked whether the subject had experienced any adverse events.

Although this is a single-arm study, participants will be masked in the sense that they will not know whether the device is operating in the active mode (intervention condition), when it gives warnings, or the silent mode (control condition), when it does not give warnings. The device will be programmed to generate a different randomized active/silent schedule for each subject. The investigator conducting the statistical analyses will also be masked as to whether the device was in the active or silent mode