Newbie but not newbie - calling all KS16s and V8 owners

[tenofnine]

In my experience with the most recent app you can program Inmotion wheels to either dip forward or tilt back a bit (or stay neutral) as the wheel senses side-to-side tilting (it depends how you rest it in standby when you are doing the F and T correction). Basically you can have it do what you prefer when turning or tilting, very customizable, not sure why you keep thinking a feature is a bug. You have full control over what it does or you can leave it alone, choice compared to no choice. The sensors are incredibly accurate and sensitive.

Then you can also manually set the pedals "neutral" position from -8 degrees to +8 degrees...it's weird before I sent mine in to get a proper controller I had to set it to -3 but now +1 is ideal. Having a constant forward tilt let's you ride tilt back comfortably but of course that is dangerous at lower battery 

And the v10 also has an additional "ride" type feature for soft/hardness

Although the app can take a few tries to connect I love the amount of customization, especially being able to change every sound it makes (and to pretty good quality too!)

[Marty Backe]

1 hour ago, meepmeepmayer said:

I'm not a native English speaker and didn't really think about the term too much So don't intepret too much into this, though I do love language discussions Google's first result is this, which fits what I meant: https://www.urbandictionary.com/define.php?term=crutch

(Maybe replace "lazyness" with "cheapness" or "winging it".)

Here:

Problem: The tilt sensors in our wheels are not very robust and get confused by a combination of a bad calibration (where the wheel was calibrated when leaning sideways) and tilting the wheel sideways (riding curves), which leads to pedal dipping in curves (mostly forwards, but the strangest things can happen, like forward on a left turn and backward on a right turn).

Ideal solution: Use better, more robust sensors.

Crutch solutions:

• Tilt back the pedals when a sideways lean is detected, which together with the unwanted dipping keeps the pedals level in curves or even gives a slight tiltback.
• Pretend the dipping is a feature aka "corner assist".

The secret is out. You've been using Google Translate all these years. I knew it!

[meepmeepmayer]

17 minutes ago, tenofnine said:

Basically you can have it do what you prefer when turning or tilting, very customizable, not sure why you keep thinking a feature is a bug. You have full control over what it does or you can leave it alone, choice compared to no choice.

• That's a great idea and nice feature I didn't know about.
• It's not really what I'm talking about.

What one would think happen: The wheel is always level and adds some forward or backward pedal tilt in curves as set in the app.

What really might happen (or not, who knows): The pedal dips forward in turns because the shitty sensor thinks that this level while it is not. To mitigate this, the wheel is set to apply "tiltback" (from the wheel's view) in curves so the pedals actually stay level. And only after that any extra tilt as set in the app is applied.

I don't trust EUC manufacturers enough to rule out the second option. Or at least those manufacturers where any pedal dipping behavior can be influenced by the calibration only, indicating the dipping was unintended to begin with. And any wheels that show the same dipping behavior, whether a calibration fixes it or not.

[meepmeepmayer]

2 hours ago, meepmeepmayer said:

I'm not a native English speaker

10 minutes ago, Marty Backe said:

The secret is out. You've been using Google Translate all these years. I knew it!

Sorry, I don't get it, that is a non-sequitur for me

[esaj]

27 minutes ago, meepmeepmayer said:

• That's a great idea and nice feature I didn't know about.
• It's not really what I'm talking about.

What one would think happen: The wheel is always level and adds some forward or backward pedal tilt in curves as set in the app.

What really might happen (or not, who knows): The pedal dips forward in turns because the shitty sensor thinks that this level while it is not. To mitigate this, the wheel is set to apply "tiltback" (from the wheel's view) in curves so the pedals actually stay level. And only after that any extra tilt as set in the app is applied.

I don't trust EUC manufacturers enough to rule out the second option. Or at least those manufacturers where any pedal dipping behavior can be influenced by the calibration only, indicating the dipping was unintended to begin with. And any wheels that show the same dipping behavior, whether a calibration fixes it or not.

While the sensors are far away from perfect, if they were really shitty, we wouldn't stay upright in the first place. Actually, they can be quite accurate, but from what I've gathered and learned when making the self-balancing robot, it requires so-called sensor-fusion, that is, using both the gyro (rotation) and accelerometer (change in position) -information and then applying some filtering on the data to produce values for the PID-loop. The "inverse pendulum" is an age old example (I mean it, control theory classes seem to have used it as a an example at least for several decades, just the sensors weren't available on the cheap until this millenia and/or no-one just figured a use case before the first prototypes last decade + Solowheel early this decade). As for the actual math behind it, I can't comment on that, I can't even really comprehend the equations, you need an automation engineer / physicist / mathematician for that 

I've thought about the pedal dipping before (the FW did it), but never came to a conclusion of what the actual reason is. If the wheel logics follow a single axis only (the "pitch", up/down tilt over sideways axis), then assuming the hardware values aren't (much) affected by sideways tilt (roll) or yaw (they shouldn't even sense that, unless it's a "9-axis IMU" with a magnetometer), the only reason for such behavior could be software. But since I don't know how the wheels really do it, there might be more to it (like filtering out the effect of the other axis).

Any control theory / automation eng or such around? I'd be interested to know, although if you drop me with a Wikipedia-page full of equations, I have my doubts how much that will actually help...  

[tenofnine]

19 minutes ago, esaj said:

While the sensors are far away from perfect, if they were really shitty, we wouldn't stay upright in the first place. Actually, they can be quite accurate, but from what I've gathered and learned when making the self-balancing robot, it requires so-called sensor-fusion, that is, using both the gyro (rotation) and accelerometer (change in position) -information and then applying some filtering on the data to produce values for the PID-loop. The "inverse pendulum" is an age old example (I mean it, control theory classes seem to have used it as a an example at least for several decades, just the sensors weren't available on the cheap until this millenia and/or no-one just figured a use case before the first prototypes last decade + Solowheel early this decade). As for the actual math behind it, I can't comment on that, I can't even really comprehend the equations, you need an automation engineer / physicist / mathematician for that 

I've thought about the pedal dipping before (the FW did it), but never came to a conclusion of what the actual reason is. If the wheel logics follow a single axis only (the "pitch", up/down tilt over sideways axis), then assuming the hardware values aren't (much) affected by sideways tilt (roll) or yaw (they shouldn't even sense that, unless it's a "9-axis IMU" with a magnetometer), the only reason for such behavior could be software. But since I don't know how the wheels really do it, there might be more to it (like filtering out the effect of the other axis).

Any control theory / automation eng or such around? I'd be interested to know, although if you drop me with a Wikipedia-page full of equations, I have my doubts how much that will actually help...  

There is a lot of beautiful fuzzy logic that is inside the programming of these devices. Leaps and bounds more impressive than anything inside a Segway (even current models) and any hoverboards ever made. I am especially impressed with Inmotions sensors and programming, but again I've only ridden 3 companies....

I think the pedal dip is caused by a simple confusion in bad programming, not the sensons at all (I agree with you on this). Usually when you go for a sharp turn or a turn at higher speeds you naturally slow down a bit, so the pedals dip forward a bit to accommodate the anticipated change in CoG  while decelerating. This is further exacerbated by the fact that while turning the pedals make an even more exaggerated angle to the ground because of the physics of riding on one wheel and nothing limiting your movement across the 2 free axis. A perfectly flat pedal in relation to the wheel can be a 10-20 degree dip in relation to the ground while turning and tilting the wheel.

I did my senior design EE project on signal and systems fuzzy logic to explain the programming behind things like the awesome Ballbot, but haven't done anything directly related to balancing systems of PEVs (so take what I say with a grain of salt just theory-crafting)
 

[meepmeepmayer]

55 minutes ago, esaj said:

I've thought about the pedal dipping before (the FW did it), but never came to a conclusion of what the actual reason is.

28 minutes ago, tenofnine said:

so the pedals dip forward a bit to accommodate the anticipated change in CoG  while decelerating.

No reason. No.

Wheel with pedal dipping -> new calibration -> no pedal dipping. Happened often enough.

The only conclusion I can draw from this: The pedal dipping was never intentional or designed in the first place. It must be a weakness of the tech. It's a bug, not a feature. Otherwise, the wheel behavior wouldn't just change from doing a calibration.

The problem seems to be that a badly calibrated sensor (calibrated with the wheel tilted sideways) seems to lose the level when it's tilted sideways again (riding curves) and thinks dipping/forward pitch is level.

55 minutes ago, esaj said:

assuming the hardware values aren't (much) affected by sideways tilt (roll) or yaw (they shouldn't even sense that, unless it's a "9-axis IMU" with a magnetometer)

No luck with that

A non-shitty sensor would be more robust in that respect. Ideally, it should be completely able to separate pitch and roll. Sensor fusion is just another mitigation for the sensor's limitations.

[Planemo]

8 hours ago, esaj said:

While the sensors are far away from perfect, if they were really shitty, we wouldn't stay upright in the first place. Actually, they can be quite accurate, but from what I've gathered and learned when making the self-balancing robot, it requires so-called sensor-fusion, that is, using both the gyro (rotation) and accelerometer (change in position) -information and then applying some filtering on the data to produce values for the PID-loop. ...  

I think you are on the right idea here esaj. I have added a few of my thoughts over on the similar Z pedals dipping thread

[erk1024]

12 hours ago, esaj said:

it requires so-called sensor-fusion, that is, using both the gyro (rotation) and accelerometer (change in position) -information and then applying some filtering on the data to produce values

It's an interesting problem. The wheel has to figure out what its orientation is. That easy enough when it's stationary, you just have an accelerometer and gravity will pull it in the right direction. But as soon as forces start acting on the wheel, then the reading from the "down" accelerometer gets modified. So that's where you have to take data from the IMU's, gyros, etc to correct that "down" calculation. I'm guessing the wheel has a sliding window type algorithm so that it looks at data over the last series of seconds to keep track of the current orientation. The problem here is that IMU's aren't perfect and they can experience drift as errors add up. Six degree of freedom (three dimensions of position and three of rotation) devices often have this problem. Now at any point if the wheel has zero forces acting on it (no accelerations except gravity) then that "down" accelerometer can be taken at face value again and the whole system can be recalibrated. This doesn't mean the wheel has to be stationary, just moving at a constant speed.

One way to get around the drift problem, and one that's being used in drones (and probably driverless cars) is to use cameras. The cameras look in different directions, and if the image rotates, or slides, then the device knows that it's moving. With stereo cameras in a direction, depth information can be gleaned by parallax. (Although I believe the driverless cars are using LIDAR for that purpose.) So you can use that data as more input to the sensor fusion your were talking about. The problem with using cameras on a wheel is obvious. In dusty situations the image could be obscured or the cameras could become blocked by debris or water, etc. But in that case the wheel could warn the user that she needs to clean them off. 

Sadly, I don't know the gory details of any of this, and I bet those details are super complicated. I've heard that there is some open source code for the camera inside-out position and orientation tracking. 

[Planemo]

19 hours ago, meepmeepmayer said:

For now (however short or long), EU law still beats UK law.

I'm not sure where you drew that statement from.

Some EU member states will have national 'derogations' for some laws, but there certainly isn't one to allow the free use of an EUC in the UK. That said, if you know of this magical law then please let us know

[meepmeepmayer]

EU regulation 168/2013. It may or may not have an effect on the details of the UK laws and whether they are applicable. It's only about what vehicles need licenses etc. and notably does not include self-balancers and one-wheeled vehicles. So if the UK failed to have requirements for these separately and specifically, it may or may not help in some way.

That doesn't mean EUCs are free use in the UK. But certain aspects could perhaps be influenced? E.g. in Germany one could argue (nobody ever did) that driving a MPV without a license or insurance is ok with EUCs for this reason.

Unfortunately not quite the magical law. We'll have to rely on magical reality (ride anyways, legal or not) for now

[Planemo]

2 minutes ago, meepmeepmayer said:

So if the UK failed to have requirements for these separately and specifically,

This.

There is no derogation or specific law relating to euc's, unlike E-bike laws/mobility scooters etc. So euc's come under the 'catch all' law of being a Mechanically Propelled Vehicle (anything powered by steam, electric, petrol, oil, diesel or gas). The term is not actually defined in the Road Traffic Acts, it is instead a matter of fact for the court to decide, but I am 101% sure they would deem an euc as such once they know it is powered entirely by electricity. Oh and added to the fact that most do 15mph+. I can almost see a Magistrate having a sense of humour failure as I type.

I can't see us being legal on our euc's in the UK for many years. It didn't happen with the original Segways and they are arguably a lot safer. The e-bike manufacturers who want to stay legal have even had to stop selling bike kits with throttles in the UK as using one takes it outside of e-bike laws into an MPV. All we can do is keep signing petitions but I just don't think theres enough euc voices out there to be heard tbh