A cane will serve as a navigation tool and it'll warn cyclists that he is blind. These accidents tend to happen because the cyclists believe that the pedestrians are going to move to the side to let them go by.As he does not see them coming, there is no opportunity to move.

Is he connected with his local blindness organization? If he lives in the US, there should be some sort of state agency under the Department of Human Services. It sounds like he could use orientation and mobility instruction, and a white cane if he doesn't already have one.

From reading all the comments, this is a policing issue. Not necessarily the boys in blue, but possibly neighbours, community leaders or local authority wardens telling people to obey the law or at least ride responsibly.

There are a few different kinds of proximity sensors. One is the $1, another is being developed by a company called $1. I'm pretty sure I've also encountered others. You can try google searches for "wearable proximity sensor" or "navigation aids for blind and visually impaired" to see what else is out there.

I'm going to write a long answer. I develop assistive tech with sensors. Feel free to post follow-up questions, send me a direct message, or tell me "Stop! Oh please, just stop."

Your uncle uses a cane and seeing eye dog. He's already done a LOT.

vwlsmssng's comment about policing is the most straightforward at this point. Report all such accidents to the police. Attend community meetings. Push for large fines.

And thank you for the question! I hope to summarize some of what I know from working image processing, 3D sensing, machine learning, and also from discussions with O&M professionals and members of the BVI community.

There are several problems that make the sensing technology necessary an impracticality, assuming it's not simply infeasible. I'll summarize briefly first, then go into a bit more detail.


1. At bicycle speeds, a large sensing range is required.
2. Only a few sensors can provide meaningful data at great enough ranges. Those sensors are not wearables. The sensors most typically used for autonomous vehicles are nowhere near small enough or light enough to be worn.
3. Omnidirectional sensing would be required. That is, it would be necessary to sense all around, not just in front the way a sighted person sees. (This point in particular is often missed by sighted engineers.)
4. Legal liability is significant.
5. Identifying objects on roads is already hard, and would be very difficult indeed on a sidewalk.
6. Reaction time is a limiting factor. Even if a threat is detected, what could your uncle do?
7. Machine learning--techniques for sensor-based systems to automatically recognize things the way people do--sucks worse than most people realize.


**Bicycle speeds.** Even assuming a bike has slowed down somewhat to travel on a sidewalk, it could be traveling at a pace much faster than a fast walker. If the bike were traveling slower, it might stop in time, assuming the cyclist is paying attention.

**No suitable wearable sensors.** Long-distance wearable sensors don't exist. There are sensors that are arguably wearable, but not as off-the-shelf assistive technology. And the sensor alone isn't the problem; significant computing power and battery life would both be required.

**Omnidirectional sensing required.** Detecting in just one direction isn't enough. Safety-critical sensing should be, and I claim needs to be, 360 degrees. The spinning sensors that can do this are neither light nor generally suitable for wearing. Omnidirectional cameras for various reasons wouldn't be suitable for detection fast-moving objects.

**Legal liability.** This is a biggie, perhaps the biggest obstacle. Assume for the moment that a weightless omnidirectional sensing device could provide warnings well in advance. What happens in court when someone wearing the device is injured or killed? Even if the device were working perfectly, imagine the arguments that the device didn't perform as expected. But well before that, consider the significant barrier to even testing a device meant for safety. I indefinitely delayed development on a related project for this reason.

**Reaction time.** Assume that perfect device again: weightless, long sensing, omnidirectional, instant feedback. A cyclist could take a corner at speed in less than your uncle's reaction time, which we'll say is 0.75 seconds. And even then he'd have to do something when he reacts.


**Identification on a sidewalk.** Autonomous vehicles have lots of sensors, lots of computing power, lots of companies spending lots of investor money, etc., and yet I still can't buy one. And I certainly couldn't lift all that tech. And that's for something running on flat roads with other cars. Sidewalks are complex, crowded, and clustered with lots of stuff cars don't have to deal with.

**Machine learning.** At least some of the techniques required for this use would rely on machine learning. And not many people would accept machine learning algorithms that are only accurate 80% of the time. 90% if you're really lucky.


There are more issues. Feedback to the user would be hard to implement for technical reasons and human reasons; for example, haptic feedback has heaps of problems. Warning sounds have their problems. Text to speech requires time, cutting into the reaction time.


And to top it off, a lot of sighted engineers who develop assistive technology start with weird assumptions like the notion that replacing a white cane would be good, or that BVI folks should only be brought into discussion after engineering is halfway complete, etc. That's a whole 'nother story.

What percentage of the collisions are head on versus from behind?

Have you looked into the Aira app? A visual assistance agent could easily warn him about oncoming cyclists.

I don't think the solution is technology. If he's not using a cane, this is one reason to get one. If I don't have my cane out, even if I'm being guided by someone else, I constantly get pummelled. It's not really about knowing people are coming, it's about them understanding you're really not going to slide out of the way, at least unless you can hear them, which you often can't on a noisy street. No, I don't like having to be a walking signpost to the situation I'm in, but I'll take that rather than a pedal cycle to the face.

Might a hi-vis vest also help?

I’m talking to some other guy randomly about a sonar haptic wrist band he’s made. I can try to find the link if you’re interested.