Robotic platform

[dakota77]

Hello

I’m building a robotic platform using R NET controller joystick and 2 motors coming from a Puma 40 wheelchair.
The robot is driven with a micro computer and an Arduino Mega driving 2 servos .
These 2 servos activate the R Net joystick.
I would like to tune the parameters.
Do you know where I can find the R Net dongle ?
Or perhaps is there another method to access to parameters menu ?

Thanks for your help.

[Burgerman]

Any high street dealer can sell you one. You just need to find a half decent one. And then PM me for the OEM version software that you will need.

Sunrise medicals price for e.g is around 260 uk pounds.

These 2 servos activate the R Net joystick.

Thats an awful solution.

[Burgerman]

Its possible that the chair has had OBP enabled. That allows dealer/end user level programming with nothing but he joystick. Its not very useful for what you are doing, nut may be useful. To enable this requires the OEM level access and it doesent work properly on any permobil with a gyro.

[dakota77]

Thanks for your reply

I have tried to hold down the buttons power and honk to enter into OBP mode but without success.

On Ebay this dongle seems to be be very expensive. (499 $)
The main things I would like to change :
reverse : the main wheels in front of the robot. (currently the main wheels are at the rear)
reverse speed : when the stick is pushed backward, the speed is very low. I would like to increase.

Thanks

[Burgerman]

I think you may underestimate the sheer volume of settings. Theres around 1000 or more. PM me ad I will give you a link to the software. So you can open and take a look.

That dongle is around double the full new retail price...

[slomobile]

As someone that has personally built arduino based robots from power wheelchairs, using Rnet for a robotics platform is probably not a rabbit hole you want to go down without very good reason. Rnet is easy to control for people. It is not easy to control for microcontrollers.

It does not have closed loop control. Open loop only unless you find the rare Rnet Encoder module, and encoders, and can somehow figure out how to use it.
If you figure that one out, please let me know, I haven't.
Rnet does not really know how fast its motors are turning. It estimates, and the accuracy of those estimates vary a lot. It is very hard to program an accurate turn in varying terrain because the motors will wind up and run on longer in rough terrain. So you need to use a gyro to modify your inputs because you do not have direct access to the outputs. This makes for a slow response and overshoot. Closed loop control cannot reliably be added on top of Rnet because Rnet is not really deterministic.

If all you want to do is make a teleop (remotely drivable by a human observer) Rnet machine, that isn't too difficult. Avoid the dongle, use CAN2Rnet, or a few other ways I can show you if you are interested.

But if you plan to do anything that requires autonomous driving and Rnet dongle, there are motor controllers far better for closed loop robotics that are cheaper than the dongle and will interface easily over serial, or PWM, documented CAN, SMbus, etc...

[Burgerman]

So you need to use a gyro to modify your inputs because you do not have direct access to the outputs. This makes for a slow response and overshoot. Closed loop control cannot reliably be added on top of Rnet because Rnet is not really deterministic.

Thats VERY true on most chairs. But its much much more accurate and responds extremely predictably without overshoot or slowing at all under a load once you properly configure motor load compensation. And turn all the important parameters up to 100%. Thn it does exactly what its told with just enough delay to make it controllable by hand.

But for automated robotics use I would still use the roboteq.

[ex-Gooserider]

I would agree with the others about not using R-net as a control system for a project like this, but if you do, I would definitely not use servos to control the joystick...

There are other threads here about using an RC controller to remote control chairs, and the basic approach is to replace the joystick (not the pod, just the actual physical stick) with digipots that are wired to emulate the analog voltage swing put out by the physical joystick... This is easily controlled by Arduino type micro controllers and will avoid all the complications of trying to physically link servo's to a stick. It also gets rid of all the slop / backlash that is inherent in a mechanical linkage, since the Arduino sends precise control signals to the digipots....

ex-Gooserider

[dakota77]

Hello
Thanks for your comments.
I understand that using servos to control the joystick is not the best way.
But for me, up to now, it does the job I want it to do.
I have built several other smaller robots using Sabertooth controller connected over serial to an Arduino Mega.
So of course I would like to remove RNET control unit and joystick and replace them by a motor controller.

Do you know some models that would be compatible with these wheelchair motors ?

Thank you.

[Burgerman]

You neeed something that can cope with a very low impedance as these will try to draw around 300 amps or so at stall or if you apply power rapidly. So it must sense current and limit it. If you want any sort of sensible control it must also allow you to use some code that adds load compensation and tank mix style steering with proper logic or things get very messy fast. Been there got the T shirt.

I used a Robotic controller, the 150A per channel Roboteq 2450 in y chair. Without a years development and testing with Lenny on this forum writing the code it would never have been usable.

An example of the sort of thing I mean is this:
https://www.wheelchairdriver.com/BM3-co ... Script.txt

Theres some pics of what this really involved here:
https://www.wheelchairdriver.com/BM3-construction/

[slomobile]

load compensation and tank mix style steering

Load compensation is unnecessary if you have position feedback and program control. Definitely good to have if you don't have encoders/odometry, but it isn't a common term in robotics circles. I've had little success finding it specified in robotics controllers.
"tank mix style steering". Tank style is controlling individual L/R wheel speeds directly, mix is converting throttle and steering inputs into wheel speeds.
Robots usually use both. Low level uses tank style odometry, high level motion planning uses steering and throttle abstractions. Kind of like humans.

[LROBBINS]

Roboteq firmware does not have motor compensation. It was something added in the script we use. The Roboteq firmware also doesn't handle acceleration correctly if you are changing both speed and direction - e.g. if both slowing and stopping turn, it changes turn rate and speed simultaneously with the result that the chair has continued to turn to a new and unintended heading. That too is corrected in the script by giving priority to changing the turn rate before applying full speed acceleration/deceleration.

[Burgerman]

This applies to every robotics controller that I checked out before buying the roboteq. Its why when you watch robotwars they spend a lot of time spinning about and trying to hit their target and missing. They all have the same issues.

This stuff is all built into mobility controllers to ensure accurate predictable control once you disable all the rediculous turn acceleration and deceleration in the programming options.

Using a robotic controller seems simple and thousands do it. But they are all pretty much terrible to control. Without something like that script that you should really examine above. Because its like night and day. Took hundreds of revisions and much testing over a very long time to get right.

Best read the pinned monster roboteq thread. If you plan on doing this in any sort of sensible fashion.

The deeper you get into it the more you will realise quite how much you didnt know that you didnt know!