making a I2 stiff ......

[quade]

Originally Posted by budsiskos

well you would have to dissasemble the mounting bracket that the leansteer pivots on and find if it uses a potentiometer or a rotary encoder.

It's a pot. Nothing even all that high tech, just a precision one. I see no reason why it could not be moved to a rigid handlebar.

[Bob.Kerns]

Originally Posted by quade

It's a pot. Nothing even all that high tech, just a precision one. I see no reason why it could not be moved to a rigid handlebar.

Unfortunately, it's a pot connected to a CPU that has software that expects the pot to correspond to a shaft that is aligned with the local vertical -- which it won't be, attached to a handlebar.

Now, you could take some rate gyros and accelerometers, a CPU, and figure out where local vertical is, and integrate this information with the info from the pot, and use that to drive an MOSFET bridge to stand-in for the original pot connection...

[quade]

Originally Posted by Bob.Kerns

Unfortunately, it's a pot connected to a CPU that has software that expects the pot to correspond to a shaft that is aligned with the local vertical -- which it won't be, attached to a handlebar.

I'm not really seeing how it would care one way or the other; it's not really part of the balance equation, just, if I haven't missed my guess, part of the bias in the torque between the two wheels.

It certainly doesn't care at all in straight & level.
It -might- care slightly, if riding perpendicular to an incline, but again, I see no reason at all why the operator couldn't adjust for that.

[Bob.Kerns]

Originally Posted by quade

I'm not really seeing how it would care one way or the other; it's not really part of the balance equation, just, if I haven't missed my guess, part of the bias in the torque between the two wheels.

It certainly doesn't care at all in straight & level.
It -might- care slightly, if riding perpendicular to an incline, but again, I see no reason at all why the operator couldn't adjust for that.

True about straight & level.

More than slightly, if there's any lateral slope. But could an operator compensate? Interesting question, best answered with experimentation.

A twist grip wouldn't be the only option, either -- other types of operator interaction might make it easier or harder for the operator to compensate.

It's definitely going to behave differently than a Gen 1 on any cross-slope. I suspect it would be difficult to control, but humans ARE pretty adaptable.

[Gihgehls]

It would be pretty unsafe. There was a thread where I was explaining that it is best to take bumps at an angle which allows the base to roll side to side with one wheel climbing at a time. A speed bump is the most extreme example, but we probably go over many small bumps all the time. When we do this on a gen2, the software is constantly compensating for the bumps by NOT turning. If a rigid mod were to be used, the operator would have to respond INSTANTLY else the machine would turn. Imagine if you had to manually adjust the torque output of the motors when you went over bumps. As it is now every tiny little imperfection in the road is considered and adjust for so the ride seems smooth, otherwise tiny bumps would cause you to trip.

To put it simply, the steering control is far more than just a differential between the speeds of the wheels. Inc calls it "roll compensation" When one wheel goes over a bump that wheel needs to travel further. I suppse a human operator could attempt it, but he/she would need ninja-like reflexes to respond as quickly as the control boards do. I'm with bob in that for this to work, there would have to be a software hack somewhere

[KSagal]

Originally Posted by Gihgehls

It would be pretty unsafe. There was a thread where I was explaining that it is best to take bumps at an angle which allows the base to roll side to side with one wheel climbing at a time. A speed bump is the most extreme example, but we probably go over many small bumps all the time. When we do this on a gen2, the software is constantly compensating for the bumps by NOT turning. If a rigid mod were to be used, the operator would have to respond INSTANTLY else the machine would turn. Imagine if you had to manually adjust the torque output of the motors when you went over bumps. As it is now every tiny little imperfection in the road is considered and adjust for so the ride seems smooth, otherwise tiny bumps would cause you to trip.

To put it simply, the steering control is far more than just a differential between the speeds of the wheels. Inc calls it "roll compensation" When one wheel goes over a bump that wheel needs to travel further. I suppse a human operator could attempt it, but he/she would need ninja-like reflexes to respond as quickly as the control boards do. I'm with bob in that for this to work, there would have to be a software hack somewhere

Wow, I am pleased to say that I completely agree...

One of the problems with oversimplifying the different processes in the segway, (and I have done it) you pretty much always wind up with a machine that no longer works, or at least no longer works well...

To people who see a segway in use for the first time, the fact that it exists at all, and it works, is like magic. It is looks too simple to work.

To people who see the innards and start to understand the way it works, each of the simple things it does are incredibly difficult to duplicate as well...

The segway does not do anything that special, and in many ways can be duplicated using simpler systems, but each would take away a percentage of this safety margin, or a percentage of that compensation factor, and very shortly, you wind up with a machine that only works on paper...

For me the uniqueness of the segway is not what it does, nor how it does it, but that it does it so well, that almost all levels of people start to take all that it does for granted, and shortly stop to see it at all...

That is the best magic of all. For those who know what I mean, the machine is magic because it disappears in use. It is, it does, but mostly it becomes not really a device at all, just part of us...

[Bob.Kerns]

Originally Posted by Gihgehls

To put it simply, the steering control is far more than just a differential between the speeds of the wheels. Inc calls it "roll compensation" When one wheel goes over a bump that wheel needs to travel further. I suppse a human operator could attempt it, but he/she would need ninja-like reflexes to respond as quickly as the control boards do. I'm with bob in that for this to work, there would have to be a software hack somewhere

That said -- I was leaving something unstated in my later reply where I acknowledged that trying it and seeing what people could do would be informative. I should have been more explicit:

For every device, there is a safety envelope. Screw drivers, spacecraft, and Segways.

There is no doubt that this would shrink the safety envelop. That doesn't necessarily render it useless, however. If people can perform well enough across a wide enough range to make it useful for people who could not otherwise make use of it, it's worth considering.

You might find the infokey needs to be set down to 4 MPH and really crank down the maximum turn rate, or something -- but while you or I might find that horribly sluggish, the users we're talking about needing this are going to find that liberating! Perhaps not as nice as it is for us, but if it's the best they can manage...

Or it might be horribly unsafe on anything but a dead-level surface. Probably not quite that bad, but perhaps bad enough to make it a bad idea.

But I won't rule it out entirely, as a limited mobility aid.

As an alternate steering mechanism for general outdoor use at 12.5 MPH, however, I think we can rule that out. There is no way it could be used at full speed over the stuff I routinely go over, and every time I do, I marvel at the ingenuity and sheer effectiveness of the leansteer system.

A gen1 can manage, though not quite as well, because it doesn't have this built-in tendency to head downhill without differential input. It basically assumes it's on the flat and level, and should go straight if the steering input isn't changed, and usually that's the user's intention.

But a steep curb cut with the street on your left would now mean, "turn left now, hard!", because it's that difference between vertical and the LSF that's the steering input.

[Gihgehls]

Gen1 might do it's own sort of roll compensation. If you get one square with a curb or parking stone, but offset it so only one wheel is aimed at the obstacle and the other has a clear path, then move straight forward, even though one wheel has to travel a longer distance, the seg will still be square after passing the obstacle. Man that was a lot of commas, sorry.

[Bob.Kerns]

Originally Posted by Gihgehls

Gen1 might do it's own sort of roll compensation. If you get one square with a curb or parking stone, but offset it so only one wheel is aimed at the obstacle and the other has a clear path, then move straight forward, even though one wheel has to travel a longer distance, the seg will still be square after passing the obstacle. Man that was a lot of commas, sorry.

That's OK, they're little, and don't use up much pixel dust.

That's not "roll compensation", it's "yaw correction". In the absence of a steering input, if anything leads the Segway to yaw from a straight line, it will seek to correct.

In the presence of steering input, if anything leads it to deviate from the path indicated by the steering input, it will seek to correct.

I don't know if there's a role for roll compensation in steering (other than in computing the steering input from the LSF position). The roll data might possibly be helpful in predicting yaw deviations, allowing a degree of predictive control. I don't know. Predictive control is difficult to do well -- if the prediction is wrong it can make things worse, and can lead to instability.

[Gihgehls]

Well, I did say a sort of roll compensation, but you're right, it is more related to yaw, although I remember asking about it at segfest 06 and one of the software guys told me that the gyros can give them yaw data but they don't use it. I think all the correction is done in a sort of dead reckoning style with the motors. I think they call it roll compensation on the i2 because they had the idea to tilt the entire control shaft at the base in order to turn, but when the base rolls, that needs to be compensated for. Something I found interesting is that they didn't change out the procs on the CU boards with gen2, they just added an extra processor in the console, the sole purpose of which is to "massage" or pre-processes the steering input before it gets sent to the CU boards. The roll compensation is done on a curved scale to keep the seg controllable on a steep side slope, and the curve is different on turtle mode as of the last update.

One fun thing you can do to demonstrate the "yaw correction" (easier on a gen1) is stand in front of it while it is in balance mode and try to forcibly yaw it with your hands by pushing on the fenders or trying to turn the wheels. If you're strong enough you can get it to rotate several degress, and when you let go it will gradually return to where it was. One of the things that I've always noticed is that the seg keeps track of where it wants to be and where it actually is, which is nothing special. But what is special is how it situationally determines how quickly to attempt to transition between those states. When you start to feel it, it's almost as if the segway is controlled by springs. That probably doesn't seem like a good analogy but I don't want to clog up this thread too much explaining it. Feel free to PM me if you want to hear more.