Is Motorcycle Blind Spot Detection Practical

I'm a motorcycle-riding Arduino newbie. A problem which has been getting worse since Covid is that people don't see or actively ignore motorcycles and motorcycle riders.

I would like to build a warning system to keep me aware of what's going on behind me. I've looked over the several posts and have not found any success stories. Sometimes the questions sound poorly though out even to me. Other times there were inherent hardware issues that the project originator either overlooked or optimistically ignored.

I'd like some informed opinion if the project is feasible before I spend a lot of time. I don't need someone to do the work for me, just point out the obvious (to them) problems I'm missing.

The failed attempts used ultrasonic detection. This has several problems including Doppler effects, short range of the sensors, interference from wind, road noise and rain, and inability of the hardware to differentiate between targets. I suspect there are also problems distinguishing between a moving object closing on you and a stationary object receding.

The LiDAR example I found was discussing 3D mapping from an aircraft. This implies that the range issues can be solved (aircraft operate at +2,000 feet generally, I'm interested in 200 feet tops and vitally interested in the nearest 50 feet or so) .

There was also mention of a low power microwave sensor but the ones I've found have a roughly 7 meter (21 feet) range. By the time the object is that close it is too late to do anything.

I would expect to use 2 or 3 sensors. One pointed forward to establish ground speed and one or two pointed backward to detect net movement toward the rider. With two sensors I may be able to distinguish targets closing on the right or left side.

Objects in a 190 degree arc centered behind the rider are the most interesting. Once the object is clearly in view of the rider detection and evaluation of the movement is useful but not vital.

The initial selection algorithm would ignore targets moving away from the rider (generally stationary objects) because cars that are not approaching are not an issue.

Objects approaching from the front will fall into one of several categories. Stationary objects will have very similar speeds clustered around the speed of the motorcycle. Moving objects will either have a speed roughly double the motorcycle speed (rider speed + approaching vehicle speed) or will be nearer to zero as the rider is either catching up, pulling away, or keeping pace with the other vehicle.

An alert is issued when an object is closing on the rider from the rear. A very fast object far away is almost as important as a fairly slow or relatively stationary (moving at the roughly the same speed as the rider) object near to the rider. An alarm is issued for a fast object near the rider.

Having been made aware of a potential problem the rider can check mirrors, look around, panic or take what ever action seems appropriate at the moment.

My question is can this be reasonably done?

Will the LiDAR correctly and reliably detect moving objects in this proposed use? Possible problems include vibration (motorcycles shake a lot relative to cars), exposure to outdoor conditions (sun, rain, heat, cold) and other things I haven't considered yet. Suitable weather-proof shock isolated packaging will be provided but some things are much more "finicky" than others.

Are there software libraries available to recognize and separate moving targets using the LiDAR data?

Can this be done in real-enough time to be useful?

Thanks for your time and knowledge in responding. If it appears that the project is feasible I'll open a new thread to discuss design specifics and trade offs as I progress.

Radar detectors are now common and can be pretty cheap. Ones I've looked at for bicycles can detect a car approach from 200 m behind, for less than USD 100. Unlike Lidar, they are not seriously affected by rain or fog.

This is the first one that popped up in a search for "motorcycle blind spot radar", and I imagine there are many more:
https://www.revzilla.com/motorcycle/innovv-thirdeye-blind-spot-monitor

Kawasaki has it on their SX models. Called "ARAS".

From my perspective, it would only be feasible to use an existing solution.
There are so many different silhouettes (car, car with trailer, truck, bikes, etc.) or possibilities to missinterpret distances and movements, that you´d need thousands of hours driving and optimization.

Not sure if it would be of help but this Youtuber may be worth a look as he goes in to some detail about using radar etc: https://www.youtube.com/@jonkraft

Thank you all for your suggestions and comments. You have convinced me that this is a blind alley so I'll close the thread.

I was not aware of the existence of blind spot detection for bicycles and will do my diligence rather than try on my own.

Particular thanks to @jremington foir the Revzilla tag.

Regards, Glenn

1 Like

Whatever you find, be sure they have been tested in areas with self-driving cars. Those sensors may blind your blind spot detection.

Advantages of a group approach -- Thanks Paul, I hadn't considered car automation as a problem. You are right that there could be both types of interference -- they blind me and I blind them.

Thanks for the idea.
Regards, Glenn

And those are some of the legal ramifications of your project!

This topic was automatically closed 180 days after the last reply. New replies are no longer allowed.