Education Project for Eletrical Engineering students

[LeonAbelmann]

Dear Forum,

Thank you for maintaing such a nice community. At Delft University of Technology in the Netherlands we are currently designing a course for Electrical Engineers (2n year of bachelor program). In the course, the students work on a project where they will control an electric wheelchair using a biosignal (for muscle or brain for instance). For this project, I would like to buy a second hand electric wheelchair that I can modify. Since these students will focus on data-aqcuisition and data analysis, the mechanical part should be relatively simple. Preferably I would like to use DC motors. Can you give advise?

Friendly greetings, Leon

[Burgerman]

I suggest ebay. Almost all proper full sized powerchairs use DC motors. These have a stall current of typically 2 or 300A each. A typical wheelchair controller is between 70 and 120A limited by software and programming. If you intend designing your own control system then best of luck as its extremely complex if you intend it to be in any way safe and controllable. So your best bet is to interface with a stock system. So be sure that you can get a powerchair programmer preferably an OEM level one for whatever chair you choose. Or developing a system will be as complex as your new brain control project.

Many good control systems such as r-net have special input modules available that allows DC voltage swing to control the chair via aftermarket joysticks or alternative controls. Such as the Omni module. Making it pretty uch plug and play for your purposes. You can do the same by duplicating the actual joystick output in any normal joystick pod.

[slomobile]

If you intend designing your own control system then best of luck as its extremely complex if you intend it to be in any way safe and controllable.

There is some truth to that, I just wanted to note that the type of complexity is not the typical "hard for average joe, easy for engineer" that you probably tune out.

The complexity is in the vastly differing cognitive and physical capabilities of individual users, medical industry regulations, insurance practices, archaic protected market stagnated development, aggravated by closer regulatory scrutiny of any departures from preexisting solutions. The result has been wheelchair motor control systems far, far behind the state of the art in robotics.

Even the "best" mainstream wheelchair system, Rnet by Curtiss-Wright does not use encoders for closed loop control or allow command of individual wheel position or velocity. If those features are important for your project (very likely) choose a robotics controller. And a driver to release the electric parking brakes.
Or undertake designing something better which we would love to use and tinker with.

https://www.ampflow.com/controllers/
https://www.roboteq.com/products/produc ... 259-detail
https://www.dimensionengineering.com/pr ... rtooth2x60
https://www.revrobotics.com
https://store.ctr-electronics.com

[Burgerman]

The problem there is that it will take a couple of years to proprly develop a control system befre they can start of the project they are really interested in. And its just easier to use say a complete r-net powerchair with an IO unit or a Omni device designed to accept external dc voltage swing input for alternate control Since that already exists with a decade of development work aleady done.

[slomobile]

The problem there is that it will take a couple of years to proprly develop a control system befre they can start of the project they are really interested in. And its just easier to use say a complete r-net powerchair with an IO unit or a Omni device designed to accept external dc voltage swing input for alternate control Since that already exists with a decade of development work aleady done.

That is simply not true. An off the shelf robotics controller can be up and running on day 1. It may not have the default joystick response you prefer but A) joystick control is not the goal of this project B) joystick response is tuned in software.

DC voltage swing is inconvenient control method when the source of drive commands is digital. Scaling and response shaping is best done digitally now, not with analog any longer. Rnet can only accept steering and throttle commands which do not produce deterministic wheel speed outputs. It cannot accept individual Left/Right wheel speed commands making precise control impossible with Rnet. That fact is not obvious! Your advice can fool people into an expensive boondoggle because they assume "DC voltage swing" is proportional to motor output like every other motor controller. With Rnet, it is not.
I have fallen into that trap. It sucks. Stop sending people there.

[slomobile]

If you do choose to use Rnet with an Omni controller, here is the manual you will need https://www.gavinphilips.com/docs/pgdt_omni_manual.pdf
page 130.
1.2V swing for full forward to full reverse, or full left turn to full right turn.

Please check out the rest of Gavin Philips website which may be helpful.

[LROBBINS]

At a minimum, if using a robotics controller one will have to program motor compensation or a constant speed algorithm and most likely some kind of acceleration-priority or acceleration mixing routine so that when changing both heading and speed you get predictable heading.

[Burgerman]

I understand how complex it is. NOW. Having been there with lenny figuring it all out. You wont understand any of this unless you have. And it seems like a simple process. Yes its easy to just plug in the roboteq and drive on day 1. Several years later we were close to a usable chair. But it is not close to being as usable as a typical mobility system on day 1 or even year 1. That took and is still under development by lenny many years. And this is just the basic a to b movement. When you look at the complexities of something like the r-net system, then it must be phone books worth of endlessely rewritten and bug fixed code over many years wih huge testing and development time.

But they are not wanting to develop a powerchair control system here. Even a basic one. But to interface with one in order to continue their project of brain control. Which being lazy I would personally be interested in for a variety of other projects. Brain controlled doors, lights, heating, typing, etc would all be marvelous!

[LROBBINS]

Brain controlled doors, lights, heating, typing, etc would all be marvelous!

I'll buy that, and even brain controlled speech as long as there's a backup for when it isn't working. BUT a brain-controlled wheelchair (or any wheelchair control that depends on a PC operating system, or, heaven forbid, the cloud ---- NO. The safety issues with that are scary. Consider faults in two circumstances: going when you should be stopped (e.g. waiting for traffic to clear before crossing a street), and stopping when you should be going (e.g. half way across an intersection).

[Burgerman]

Changing he subject very slightly, I am my own "cloud". Cloud is not needed. And more importantly every damned thing ou do or buy wants you to "sign in" or log in. Non of my computers, or myself use or need any of it. I don have or want a google, microsoft or any other online account. The ONLY thing I ever sign into are a few places like reddit or github. Or here. Windows doesent log in. Or ask me to do so. I have all the "free" adobe and other sysems (like coral, office, etc) cloud based stuff installed on my PCs. But every one of those is "fixed" so that it cannot do or ask to go online to send data or check for updates etc. And all of them are blocked in the hosts file as well as in the windows firewall.

I have a list a mile long!

Yet everything works just great.

I never go online without a GOOD VPN, and I never pay for anything including paywall sites, movies or software.... Except the VPN. And I have my own "cloud" that goes everywhere with me. All work, software, backups etc live in multiple places. I can log onto my desktop PCs (wake on lan, magic packet etc) from my laptops. I can see all my files from the pub or wherever. And I can backup each computer to 3 others and use remote desktop. Did you know that you can control computer A from computer B? And in turn control computer B from the remote connection at the same time? I sometimes forget which computer is controlling which other one... I have at least 4TB disks in all my laptops, some have 2. And my desktop has 2x 2TB and 2x 4TB disks. Not exactly short of space.

What is the "cloud" actually meant to achieve? It really just data mining and sales in disguise. Just like the webmail that people think is eMail... Cloud is another name for £££ extraction or control or ads or data mining.

For e.g cloud based programs such as Adobe photoshop, Adobe premier, etc is sold on a monthly rental. So it costs you about £32 per month! Or 400 a year per PC. But my version is free, forever, never goes online. Same with all the rest.

[slomobile]

At a minimum, if using a robotics controller one will have to program motor compensation or a constant speed algorithm and most likely some kind of acceleration-priority or acceleration mixing routine so that when changing both heading and speed you get predictable heading.

I'm trying to wrap my head around this paper https://www.hindawi.com/journals/mpe/2013/919837/ and some like it to achieve torque vectoring with sliding mode control (not wheel sliding, the Russian control strategy) on the front axle of a 4wd chair with very long trail on the front pivot steering axle. And then somehow do it again mixing the fixed rear axle cooperatively with the front.

The front axle will be twitchy and always oversteer compared to the desired turning circle. The rear axle will understeer(slide sideways into the turn) or countersteer(pivot). Hopefully producing the desired vector and some slight automatic wiggling (underdamped control point oscillation) for traction when stuck.

I have only a very loose grasp on any of this at present, but moving forward.

[Burgerman]

Not sure I understand the point of it.

I'll buy that, and even brain controlled speech as long as there's a backup for when it isn't working. BUT a brain-controlled wheelchair (or any wheelchair control that depends on a PC operating system, or, heaven forbid, the cloud ---- NO. The safety issues with that are scary. Consider faults in two circumstances: going when you should be stopped (e.g. waiting for traffic to clear before crossing a street), and stopping when you should be going (e.g. half way across an intersection).

And while this would be marvelous I suspect it will be about as useless as the voice commands and robotic phone command systems, or Ai is at the beginning. It will likely take many decades to make this knd of thing in any way practical. And I am 63 now. So not holding y breath. But wish someething would speed up development by x100 or something. But that will take more than a few students playing at it. It will need £££££££££.

[LROBBINS]

Slomobile,

I too tried to read that article. I get what they are trying to do; get two motors to follow a variable track and load without "fighting" each other. To do that, they must be synchronized (both powering up or down at the same moment) and speed matched (neither one trying to go faster or slower than the other). They say they've achieved that, but the math and algorithms are way, way beyond me.

[Burgerman]

It seems like a excercise in maths with no purpose. Doesent a differential on the axles achieve the same thing and been sed for years? Or on 4 wheel drive another one between each axle?
And a simple torque load cell on eack motor/wheel allow a better solution that allows feedback loop to do the same thing? That allows for tyre wear, loose surfaces under a weel(s) and a few small hills/rocks/low or unequal tyre pressures or uneven surfaces too. There are better ways than attempting to do it with maths. So just an excersize for students?

[slomobile]

An exercise in drive efficiency. As you know fixed wheel 4wd wastes a lot of power in turns. 2 wheel differential drive wastes power off road dragging casters through the grass. Motors dedicated to steering don't contribute to drive. Untethered floppy steering wheels like X8 or rear steer like Meyra are weird, but closer to what I'm looking for. I wanted an off road capable machine where all ground wheels and motors contribute to both drive and steering and do not waste watts working against each other, or the chassis. Front pivot axle allows the addition of simple handlebars and/or steering footpegs to assist steering and go hands free.

[LeonAbelmann]

Thank you all for the excellent feedback, I learned a lot! You recommend either to use the external devices input on an R-Net controller, or to use general purpose motor controllers. The first option has the advantage of simplicity, the second option allows for better control. The project will be run by about 50 students each year, working in groups of 6. Some groups will be more inclined to work on the data acquisition and decision taking, other groups will be more interested in the mechatronics. The R-Net system seems to be highly modular. So I can imagine we allow the students to insert their solution somewhere in the loop. Some groups will could use the speciality control interface on the Omni2 (https://www.cw-industrial.com/en-gb/med ... -interface). Other groups could connect directly to the R-Net power modules or general purpose motor controllers.

From Gavin Philips website I understand that R-Net is used by Quicky and Permobil. The Permobil documentation seems to be more informative. In the Netherlands, where E-Bay is Marktplaats, I found a Permobil M3. Would you recommend it?

https://www.marktplaats.nl/v/diversen/r ... e-rolstoel

[Burgerman]

Permobil have many in house parts, not pure r-net. So thigs like addtional impossible to get programmers or even getting parts is a big problem... If it was me I would avoid permobil.

Look at sunrise chairs. Magic mobility, ami etc.

[woodygb]

Perhaps this...?

https://www.marktplaats.nl/v/diversen/r ... 43986d4a06

[thamesmud]

Changing he subject very slightly, I am my own "cloud". Cloud is not needed.
I completely agree with this, I'm amazed how many people use cloud based systems in home automation etc. My Home Assistant install does not require any cloud access though I can use a VPN to access it whilst away from home. You would have to be a bit soft in the head to rely on any system that needs on an internet connection to function without a local option ( in my opinion ). I'm even more concerned about the headlong dash to connect cars to the cloud, it's positively dystopian and as for drive by wire, don't get me started , pass my tin hat.

[Burgerman]

I agree. All my computers are set up with wake. So I can fire up my desktop PC for e.g. from anywhere. Via a phone card, the wifi in the pub, whatever. And I can store any amount of data. Which is regularly backed up. To 3 or 4 other computers. This cloud nonsense is all about ads, £££, suscriptions and you having to rely on some 3rd party. You are the product. And I just dont do any of it!

Every PC, efery office suite, every adobe program, every operating system, every search engne etc all keep insisting you log in... For what? So they can offer you better paid services, so they can track you and sell your data, so they can sell you stuff, so they can advertise at you more accuratelyWindows tries to get you to log on to microsoft. Why? So that they can do all of the above.

Theres ways around all of this. So nothing ever asks.

[slomobile]

IAll my computers are set up with wake.

***Breaking news ***
Burgerman is woke.

Just kidding.

[Burgerman]

I hope so! I am the opposite to that. I just got myself a ENOCH POWEL WAS RIGHT shirt...

[ex-Gooserider]

There is NO SUCH THING as "CLOUD"! It is just a short hand way of saying "OTHER PEOPLE'S COMPUTERS"

ex-Gooserider

[Burgerman]

There is NO SUCH THING as "CLOUD"!

Well there definitely is.

[Dan]

All your info is already stored in the "cloud". Name, address, phone number, pictures of you, your medical records..........
Keeping your own home storage/server can be expensive where my ISP give 50GB of cloud storage for free with end to end encryption. Apple give 5GB of free of encrypted storage.

Phones have about 250GB of storage anyway.

Having said all that I do run two raspberry pi's on their own domains and a NAS

BTW as far as I know brain control of a wheelchair today is impossible. Its going to be a long time for technology to catch up. The brain's activity is to weak to measure reliably, the workings of the brain too abstract and the skull too thick.

[Burgerman]

Phones have about 250GB of storage anyway.

My phone has zero web access, no email, and a minute built in storage measured in kilobytes...

And my laptops and desktops each have at least 4 to 25TB of photos and backups each alone!
Even my laptops have 4TB drives.

[LeonAbelmann]

Perhaps this...?

https://www.marktplaats.nl/v/diversen/r ... 43986d4a06

Thank you for thinking with me! In the meantime, I have been in contact with Curtiss-Wright. They have three solutions to interface with R-Net: a simple IO-module, their Omni2 and a development board. They recommended I search for a manufacturer based in the Netherlands. I am now considering DIETZ, e.g. https://www.marktplaats.nl/v/diversen/r ... r-advanced

I also contacted Gavin Philips. His feedback was very valuable, especially the observation that I should let my students interact with wheelchair users and professionals who tailor wheelchairs to the user's needs. He warned me that many wheelchair manufacturers protect the CAN bus against non-standard interfaces.

[Burgerman]

I should let my students interact with wheelchair users and professionals who tailor wheelchairs to the user's needs.

I dont know who gavin is but I suspect the "professionals" and end users will get you nowhere fast.

[woodygb]

@LeonAbelmann

Interfacing with a chair should be fairly simple and straight forward exercise for your students.

I would however suggest that it is a waste of time getting a chair , until you have a reliable signal source ( brain wave reading etc ) that you can use as an INPUT to any of your afore mentioned interfaces.

[LeonAbelmann]

I should let my students interact with wheelchair users and professionals who tailor wheelchairs to the user's needs.

I dont know who gavin is but I suspect the "professionals" and end users will get you nowhere fast.

Gavin Philips site was suggested to me by slomobile. http://www.gavinphilips.com. He remarked that having engineers work in isolation does not give the right results. Therefore he suggested "to learn from wheelchair users, and from the professionals who configure and customize their wheelchairs".

He suggested also equipping the chair with additional sensors to avoid collisions. https://luci.com/making-power-wheelchai ... er-thanks/. What are your thoughts about this?