Could AI provide personally customizable wobble dampening algorithms to once and for all eliminate this hazardous susceptibility?

[grampmiles]

Developing personalized wobble dampening algorithms would require expertise in vehicle dynamics, control systems, and specific knowledge about your vehicle's characteristics.

[KiwiMark]

I have been riding for ~3,300km. At first I wished my RS was not so prone to wobbles, now I know that it isn't prone to wobbles, it was the rider. In my first 300km or so I had a couple of wobble-induced crashes due to being a noob and not knowing what to do. In the last 3,000km I've had zero wobble-induced crashes due to being able to quell wobbles when I feel them beginning. I don't need to consciously think about how to quell the wobbles now, it is a subconscious muscle memory sort of thing.

I do wish I had been more patient and rode slower during my first 300km learning phase, it would have been less painful and may have involved zero wobble-induced crashes.

[mike_bike_kite]

I partially agree with what everyone says about how wobbles become less of a thing with experience but Marty, in the video, is a very experienced rider. I'm just an OK rider and I've had wobbles kick in when braking hard on a rough road surface. I'm also a software guy and I actually think you could eliminate wobbles with standard programming (no clever AI required). All you have to check for is the wheel moving rapidly from side to side to detect a wobble state. Then "all" you do is apply small amounts of braking or acceleration as the wheel is moving from side to side to deter the wobble - as the wheel is moving away from centre you'd apply a little acceleration and do the opposite as it moves toward centre position. This would only activate in a wobble situation and would turn off again once the wheel has recovered. I guess you'd want to allow the feature to be turned off in the settings as well. It would probably only work effectively during the start of the wobble rather than as you're about to fall off.

[Rawnei]

56 minutes ago, mike_bike_kite said:

I partially agree with what everyone says about how wobbles become less of a thing with experience but Marty, in the video, is a very experienced rider. I'm just an OK rider and I've had wobbles kick in when braking hard on a rough road surface. I'm also a software guy and I actually think you could eliminate wobbles with standard programming (no clever AI required). All you have to check for is the wheel moving rapidly from side to side to detect a wobble state. Then "all" you do is apply small amounts of braking or acceleration as the wheel is moving from side to side to deter the wobble - as the wheel is moving away from centre you'd apply a little acceleration and do the opposite as it moves toward centre position. This would only activate in a wobble situation and would turn off again once the wheel has recovered. I guess you'd want to allow the feature to be turned off in the settings as well. It would probably only work effectively during the start of the wobble rather than as you're about to fall off.

Just accelerate or brake a one wheeled device with a rider standing on it? Sounds a bit sketch.

[mike_bike_kite]

10 hours ago, Rawnei said:

Just accelerate or brake a one wheeled device with a rider standing on it? Sounds a bit sketch.

It doesn't have to be a huge acceleration or brake, it just has to be enough to stop the wobble getting worse. If caught early enough then the wobble could be fixed before the rider even realised they were about to go into a wobble. The wheel wouldn't feel like it's accelerating (or braking) by itself as the acceleration and braking would cancel each other out. It would feel more like a slow vibration. 

When you slam on the brakes in the wet in an ABS equipped car, nobody describes the computer controlled on/off braking as a bit sketch. The same could be said of traction control. Torque vectoring is done on some cars to improve cornering by braking individual wheels during a corner. This is done entirely by the computer but nobody describes the this as a bit sketch either.

It shouldn't be an expensive thing to add as no additional hardware would be required. The wheel already has an accelerometer to detect the movement. It's worth an experiment anyway. 

[Rawnei]

38 minutes ago, mike_bike_kite said:

It doesn't have to be a huge acceleration or brake, it just has to be enough to stop the wobble getting worse. If caught early enough then the wobble could be fixed before the rider even realised they were about to go into a wobble. The wheel wouldn't feel like it's accelerating (or braking) by itself as the acceleration and braking would cancel each other out. It would feel more like a slow vibration. 

When you slam on the brakes in the wet in an ABS equipped car, nobody describes the computer controlled on/off braking as a bit sketch. The same could be said of traction control. Torque vectoring is done on some cars to improve cornering by braking individual wheels during a corner. This is done entirely by the computer but nobody describes the this as a bit sketch either.

It shouldn't be an expensive thing to add as no additional hardware would be required. The wheel already has an accelerometer to detect the movement. It's worth an experiment anyway. 

A car and a unicycle is two completely different things.

[mike_bike_kite]

A car and a motorcycle are different things but they have very similar technology (ABS, traction control). I remember when ABS was introduced there were so many haters talking about not being in control. These days, most people wouldn't even consider a road bike without it.

I think you actually have to give some specific examples where you think the above idea won't work just because it's a unicycle. 

[Rawnei]

31 minutes ago, mike_bike_kite said:

A car and a motorcycle are different things but they have very similar technology (ABS, traction control). I remember when ABS was introduced there were so many haters talking about not being in control. These days, most people wouldn't even consider a road bike without it.

I think you actually have to give some specific examples where you think the above idea won't work just because it's a unicycle. 

No actually I think it's the other way around, it's you have to show that you have a viable idea here, that the physics involved are viable that any sort of intrusion in motor behaviour won't throw the rider off balance and make things even worse. Comparing to vehicles that doesn't even operate under the same physical rules (self-balancing vs non-self-balancing) doesn't really show proof of anything.

Either way this whole topic is still making a hen of a feather IMO, we don't need more complexity and things that can go wrong on these devices when this "problem" is something that can be addressed with practice, experience and rider behaviour.

Edited June 18, 2023 by Rawnei

[mike_bike_kite]

On 6/13/2023 at 4:24 PM, weelwoman said:

Perhaps AI could customize wobble dampening algorithms for individual wheels, tire pressure, unique muscular inputs, and so forth?  What do you tech wizards think??

In short, I think it's certainly possible and it would only require a firmware change. I think the explanation above explains fairly clearly how it would work. Obviously it would have to be tested to make sure it works unobtrusively. Whether people would want it so far appears to be questionable.  

[Wheelwoman]

Great conversation guys! Just my own uneducated opinion is in Mike_Bike's camp but let's hammer this out .... what if an ABS-like low cost solution to wobble potential is concocted .. WOW, how terrific for our burgeoning mobility phenomenon ... let's not be Ludditish about it (: ...... (oh oh, should I not have said that?? Tee hee, just trying to keep the conversation going... I really do love and respect ALL of your inputs)

[mrelwood]

Whether or not such a feature would actually work or not is one discussion, but how to make some instance make or modify a firmware to do that is a whole another. I don’t know if a suggestion like this would even reach high enough level at the major companies, after all there are probably all kinds of ideas and suggestions coming from many directions.

[mike_bike_kite]

2 hours ago, mrelwood said:

Whether or not such a feature would actually work or not is one discussion, but how to make some instance make or modify a firmware to do that is a whole another. I don’t know if a suggestion like this would even reach high enough level at the major companies, after all there are probably all kinds of ideas and suggestions coming from many directions.

I agree with you but it would be relatively easy to program a microcontroller (Arduino, Picaxe etc) and link it to an accelerometer and then see if wobbles could be detected and not confuse them with normal riding. The whole thing could fit in a small plastic box, sit on top of the wheel and cost under $50 (+time). I think you'd just want to detect for a fast repeating left right movement that steadily increases. You'd want to log how often it happens and perhaps sound a buzzer. It would be interesting to see if there are micro wobbles that turn into full blown wobbles. The real problem in testing is that many of us don't get wobbles very often.