PINNED - Roboteq Controller - developing for powerchairs

[Vitolds]

smooth acceleration and smooth braking at zero

[Burgerman]

But you don't always decelerate at 0 but use full back stick. Which changes steer to opposite direction like a switch! While traveling forwards. And when reversing your steering first of all works normally, and suddenly swaps over at about -500 reverse!

Try it. You wont understand till you try driving it...

[LROBBINS]

I'll repeat myself: Try thinking about where your NOSE is pointing and which way it is moving. Burgerman is absolutely correct.

Funny thing. The same issue came up years ago on the WCJ site and it had nothing to do with coding as it involved a "standard" chair. It took that person many exchanges of messages to understand that a wheelchair doesn't steer the same way as a car!

[Vitolds]

I do not want to think where my nose.
Left is left
right is right
wheelchair so is controlled by.
you do not have solved the problem!

[woodygb]

There is no problem to solve.

[Vitolds]

in a wheelchair there are no problems.
problems you have with your code

[Burgerman]

Its not possible to "solve" because of the way that a tank steered device operates. Your own tracked chair, same thing.

Theres no problem to solve. If you DO solve it, you made a real bad problem!!!

What vehicle is this going in???

[Vitolds]

at itself I solved the problem.
I wanted to experience your control.
you
left - go the right
right - go the left


your control is good when high speed. Only when you go forward.
would do so
100 -30 % - your control
30 - 0 % - control wheelchair

[Burgerman]

In english that means?

[LROBBINS]

No, if going backwards and turning right, the rear of the chair goes to the LEFT, but the front of the chair goes to the RIGHT. If going forwards and turning right, the rear of the chair goes to the LEFT, but the front of the chair goes to the RIGHT. THIS IS THE CORRECT WAY FOR A CHAIR TO BEHAVE. If you change the algorithm so that it doesn't do this, if you hold right stick and change from going forward to going backwards, the chair will reverse its turning direction - that's bad, bad, bad.

Go sit in any marketed power chair and feel what it does - AND DO THINK ABOUT WHERE YOUR NOSE IS MOVING - is it moving further to the right with right stick both going forwards and going backwards? If so, that is correct. Your design should do the same thing. Remember, a joystick is not pointing in a given direction, the way a car's steering wheel does things. It is saying - "turn further to the right at rate x, or turn further to the left at rate x". The angle you end up at actually involves an integration of "steering" over time; it's very different from an automobile (and more like the ailerons of an airplane).

[Vitolds]

I do not want to look behind nose.
I want to go backwards where Joystick reveals
this is the correct control

[Burgerman]

No its wrong. And opposite to every full sized skid steer vehicle, tank, Caterpillar or tracked vehicle. And opposite to every single wheelchair made. And opposite to YOUR tracked chair.

ALL of these and every wheelchair rotate ANTICLOCKWISE with LEFT stick. CLOCKWISE with right stick. As seen from above. Anything else would be undrivable!

[Vitolds]

is not tracked chair.
tracked chair everything is fine, set up of Lithium batteries, suspension, shock absorbers.

[LROBBINS]

This has nothing to do with skid steer per se, but with any vehicle with differential steering. The diagram that you think is correct has the front of the chair moving toward the right in what you call a forward right turn, but the BACK of the chair moving toward the right in what you call a backward right turn. If the back is moving toward the right, the FRONT is pointing further toward the left - AND THAT IS A LEFT TURN NOT A RIGHT TURN (even though the chair as a whole is indeed moving rightward, the pointing angle is changing toward the left). Get in your track chair and try to see what it does when you turn right while going backwards - the front of the chair will point further to the right if it is set up correctly.

[Burgerman]

When he tests his track chair, or any wheelchair. Or a full sized tank! He will understand.

[Vitolds]

I understand!
I see the difference!
Your smooth control over
I have roboteq!

[Burgerman]

is not tracked chair.

What is it?

[Vitolds]

[funkykeyboard]

Can you tell me where I get the software from for the Roboteq controller. I have had it once, but lost it

[Burgerman]

Software? Or script?

You want the newest firmware installing, and the matching roboteq software.
Then install the newest script that I am using. But its acceleration/deceleration levels and other settings etc will need figuring out as I only tested this latest version in the small bot so far. So maybe 5x lower or more. Or you will loop it/break it. But script is good.

download/file.php?id=5867

[LROBBINS]

Motor resistance value also MUST be changed from the MowBot settings, and if current sensors are used they need to be of a suitable current range for the intended motors with the SensorRange value set to match. Use the MowBot motor resistance with a wheelchair motor and I'll guarantee that you'll have an uncontrollable runaway.

[Burgerman]

100A sensors and range, set to 45mOhm was working great before I broke a motor last year. So if using similar motors start at 40.

[Vitolds]

contactor
Mercedes A 003 542 18 19 (Gruner 750H 12V/1000A)
perhaps it will be useful

[Dannyboy]

Does anyone have a schematic for what inputs are being used on the robotec with regards the joystick controller and pot?

[Burgerman]

Every roboteq, and every user is different.

But you will need to choose in the roborun software suitable inputs for:

DIGITAL OUT:
Brake (I use one for both brakes via a solid state relay)
Brake 2 if not as above
Status LED
Battery low

ANALOG IN:
Left Current sensor
Right current sensor
Joystick forw/rev
Joystick left/right
Speedpot min/max

PULSE
RC Forw/Rev
RC Left/Right
And a failsafe channel, (throttle on some receivers) to exceed guard band or to set a stop command) for some receiver types.

as a minimum. And calibrate each, configure correct settings both in roborun, and inside the script to match.

[Dannyboy]

That's great gives me a starting point. ta

[Burgerman]

Small robotic 4x4 skid steer using a smaller roboteq, current sensors, lennys script as used on my BM3 for testing, and a 10Ah 16.8V lithium battery. Great control. Tons of torque. Was partly built to test scripts. And usually has a go-pro cam on top that sends the image back to my laptop or goggles via a 5.8ghz tx. Allows me to surveil the house/garden/street at night while laid in bed. And freaking out the neighborhood cats. Can hide for hours then creep around slowly and follow one when I see it. Adequate range for around my local area.

http://www.wheelchairdriver.com/gopro/test-control.mp4

[Burgerman]

Newer latest script without current sensors.

http://www.wheelchairdriver.com/gopro/bot.mp4 250mb.

Still got good control. Both steer and backward/forwards axis are seperated and pure. And the motor load compensation works now without external current sensors.

[LROBBINS]

Looking good, but I think you can even improve on this by tuning the new parameters - see e-mail.

[Burgerman]

I am not quite so confident. Having played with every parameter for hours. I have almost no unloaded dead-band at all when the thing is lifted of the ground. 1mm of joystick and the wheels turn slowly.

But I have about 10% of stick travel before any compensation kicks in under load. And it does so very suddenly. No matter what settings. So that once on the deck, it will not zero turn with a small stick movement. As I add more stick, beyond about 6mm compensation suddenly kicks in and it turns.

It didn't do that with current sensors and was very smooth and powerful. It isn't that theres anything wrong with the script, I think. Its just inherent in the roboteq current sensing on battery input. I dont think it reads anything at low currents.

Its not bad. As you can see by the vid. Remembering that it takes well under 1A (a tiny pulse-width) to go forwards, and about 7A or more (about 45% pulse-width) to turn on the spot and skid all 4 wheels. And it does both with equal sensitivity, and smoothness. So compensation works very well. If you pick it up and slow a wheel at low speeds by hand, it really fights you. But it does feel less controlled than before at initial "zero turn" start. As there is a wide compensation "dead-band".