UK-based robotics engineer

You dont have to, but its interesting!

MAIN WEBSITE: www.wheelchairdriver.com

UK-based robotics engineer

Postby Davross » 21 Jan 2018, 13:50

Hi there, thought I'd introduce myself – my name is Kev, a.k.a. Davross (deliberately misspelled) and I'm a 48-year-old electric wheelchair user living in Southampton.

I have a severe neuromuscular disorder called SMA and have been using an electric wheelchair since the age of four… It was a BEC fireball which was one of the first children's electric wheelchairs available in the UK.

I have a Masters and Doctorate in electrical engineering with a specialism in instrumentation for surface metrology – this means I'm used to programming four and five axis robotics controllers to choreograph complex machine motions around parts such as artificial hips and knees.

My hobbies are programming, CAD/3-D printing and radio control aircraft, and I actually write a regular column for the BMFA about drones – primarily indoor micro racing drones.

My current wheelchair is an NHS provided Invacare Spectra XTR that is coming to the end of its life, and I privately own a near duplicate model as a spare. I'm contemplating upgrading this myself using some of the fantastic ideas provided on this site.

With the advent of free non-commercial use, project sharing CAD programs (ie. fusion 360), I have wondered if there is scope for members of a group like this to collaborate on a design of a custom wheelchair manufactured from predominantly off-the-shelf parts, combined with 3-D printing, microcontrollers and radio controlled grade equipment? Certainly LiFe cell, but possibly evolving the technology even further and adopting full brushless motors.

Cheers
Kev
Davross
 
Posts: 5
Joined: 21 Jan 2018, 13:37

Re: UK-based robotics engineer

Postby Burgerman » 21 Jan 2018, 16:25

Welcome! Custom for who though? The chair you want is going to be vastly different to mine! Thats the purpose of custom chairs.
User avatar
Burgerman
Site Admin
 
Posts: 65051
Joined: 27 May 2008, 21:24
Location: United Kingdom

Re: UK-based robotics engineer

Postby greybeard » 21 Jan 2018, 17:36

Burgerman wrote:Welcome! Custom for who though? The chair you want is going to be vastly different to mine! Thats the purpose of custom chairs.


Isn't that mainly to do with motor selection and mechanical adjustability to conform to the individual's physical requirements and intended use, and programming to suit the user's capabilities?

I recall that you said you hate "modular", but for a project like Davross' suggestion, using off the shelf parts and developing a standard multi-adjustable frame to hang it all on could be achievable. Especially welcome would be properly written and accessible easily programmable software that is practically nonexistent in chairs from mainstream manufacturers.

Although it probably wouldn't get off the ground in the US given their funding system, it will be interesting to read the responses of others,
greybeard
 
Posts: 1152
Joined: 28 Aug 2014, 20:15
Location: Dorset, UK

Re: UK-based robotics engineer

Postby Davross » 21 Jan 2018, 19:09

A modular system is what I had in mind, combined with parametric CAD that allows you to change key design parameters on-the-fly and have them ripple forward into the final product. BM has the overall control system and battery technology well and truly nailed, although I can see reasons where you would reduce overall battery capacity (and hence weight) for users that don't do a lot of outdoor, long-range driving.

Breaking the chair design down into its key components, you have a drive system bolted to a chassis, with a seating solution on top. Other than motor power and battery specification, most chair manufacturers get away with a one-size fits most for the chassis, and then combine it with a very adjustable seat solution. Of course, at the extremes, some users have very specialised requirements, but these are far easier to implement if you have a firm base to work from. Get away from "not invented here syndrome".

As BM says, at the moment the community is reliant on very specialised skills to write and tailor the Roboteq software. If we move the control algorithms up a level, and use a small microcontroller (probably a repurposed racing drone flight controller), you can make use of a very mature, open source solution (clean flight) with very good front-end for programming. You'll also gain the benefit of 6° of freedom motion control which could be used to limit acceleration/turn rates and provide automatic anti-tip if chair started to tilt over backwards. You also get a radio control interface as standard, and things like GPS integration.

These were my basic ideas – let's start a discussion and see if it has any legs :)

Kev
Davross
 
Posts: 5
Joined: 21 Jan 2018, 13:37

Re: UK-based robotics engineer

Postby Burgerman » 21 Jan 2018, 21:03

A modular system is what I had in mind,


Modular is compromise rather than pure design. Modular is a bad thing. Its only good for the manufacturer, or the price. Not for the design or end user.

Ask yourself why a formula 1 (no compromise ultimate solution) car is not modular.
User avatar
Burgerman
Site Admin
 
Posts: 65051
Joined: 27 May 2008, 21:24
Location: United Kingdom

Re: UK-based robotics engineer

Postby Burgerman » 21 Jan 2018, 21:11

As BM says, at the moment the community is reliant on very specialised skills to write and tailor the Roboteq software. If we move the control algorithms up a level, and use a small microcontroller (probably a repurposed racing drone flight controller), you can make use of a very mature, open source solution (clean flight) with very good front-end for programming. You'll also gain the benefit of 6° of freedom motion control which could be used to limit acceleration/turn rates and provide automatic anti-tip if chair started to tilt over backwards. You also get a radio control interface as standard, and things like GPS integration.


Thats not a bad idea. But roboteq already has RC input, and I can and have tried gyros etc. And so adding a flight controller is already simple plug and play with the existing script. However I personally neither want electronics taking charge or "helping" or to limit things. Other than what I can already choose! Gyros, stability accelerometer help just makes it worse. I want the stick directly connected to the wheels with as little delay as possible.
User avatar
Burgerman
Site Admin
 
Posts: 65051
Joined: 27 May 2008, 21:24
Location: United Kingdom

Re: UK-based robotics engineer

Postby LROBBINS » 22 Jan 2018, 00:33

I will be interested in following where your ideas lead, but I too am not at first blush enamored of your suggestions. Current commercial chairs are essentially modular - as you said, a base unit with other things like seating systems added on. The result is that multi-actuator seating systems don't actually share, but replicate, the structural elements of the base and add a lot of weight; up high where it really messes with stability. Instead of adding a gyro to patch that, we should be trying to get at the root cause, not exacerbating the problem by continuing the trend towards "modules".

I see progress, beyond what's already been demonstrated by John (Burgerman) and others here, coming from a different direction; from real engineering of integrated solutions, starting first with a re-think of structures. It's been at least 50 years since decent bicycles switched to Cr-Mo steel and later introduced alloy and carbon fiber, yet our chairs are still built from the cheapest grades of mild steel; even the bicycles sold by my local supermarket are heat-treated 7000-series alloy these days. A ready excuse had been that given the weight of the batteries and person, the frame makes a negligible dent. That, however, is even more true of a bicycle or motorcycle where the rider is a much higher fraction of the total weight, but it's clear that every little bit of weight saved makes for a more efficient, and better handling, machine. Moreover, with Li instead of Pb even that excuse goes away, and it never really worked once one added in seating systems made of the (almost) pig iron, anti-tips, leg supports and on and on. And as the batteries get lighter, the stability problem of a high CG gets ever worse. I'd like to see a chair made of proper structural materials, with as much function as possible integrated into the fewest number of pieces possible and with a stress-analysis of each piece so that there's not an ounce more weight than needed for structure and durability. I haven't the skills to actually do that, but I think that even I could come closer than what's being marketed.

Second, we do need more efficient motors, but if the current market is any guide we're actually going backwards with cheap worm-gear motors (with perhaps 70% efficiency) replacing helical gearboxes (the best of which reach 85% or higher efficiency). Brushless motors may yield some improvement, but they're certainly not a panacea. Simple brushless motors are efficient and durable if run at very high speeds, but our chairs have to work over a wide speed range from maneuvering indoors to the crazy speeds John likes. If gearless, they have to be heavy, multi-pole torque motors and even though Invacare did eventually get a workable design, as John found they still lack torque at low speeds. If geared, the efficiency of the gearbox is far more important than any difference between brushed and brushless, and the (more theoretical than demonstrated) durability of brushless motors really only comes into play at high, e.g. 9000 RPM, motor RPM. But for a wheelchair, the added reduction stages needed to go to wheel RPM from a 9000 RPM motor vs. a 3000 RPM motor will more than outweigh any efficiency advantage and probably not do much good for durability either. Which would you rather be a maintenance item - brush changes or worn out gears. Add to this that simple trapezoidal commutation will not give reasonable torque and smoothness at low RPM (it is like having a brushed-motor commutator with too few segments); you need sin-cos commutation and field-oriented control which in turn means adding an encoder or resolver (plus the Hall sensors most brushless motors already have), so the simplicity of not having brushes is not all that simple in the end. BTW - if anyone finds a geared brushless motor that can accept both an encoder (or resolver) AND a parking brake with manual release, I'd sure like to hear about it. I agree that we should be looking toward brushless motors, but I just don't have very high expectations that the transition will be easy or have a great payoff.

Third, as John has pointed out, there's nothing standing in the way of adding all the digital goodies to a Roboteq based controller, and doing so is especially trivial in the CANbus system I'm using. A multi-axis flight controller would just be another CAN node. Some ad-hoc programming was needed to get past some basic lacks in the Roboteq's intrinsic programming, but they had nothing to do with functions like radio control, gyro feedback, RC operation etc. The Roboteq is capable of handling the high currents we need. A multi-axis flight controller by itself doesn't provide that and you'd still need a robust motor controller.

That's enough blather for one night.

Ciao,
Lenny
LROBBINS
 
Posts: 5543
Joined: 27 Aug 2010, 09:36
Location: Siena, Italy

Re: UK-based robotics engineer

Postby Burgerman » 22 Jan 2018, 01:34

If gearless, they have to be heavy, multi-pole torque motors and even though Invacare did eventually get a workable design, as John found they still lack torque at low speeds.


And with typically 4x the current draw at stall/low speed at the same torque level as a geared/brushed motor. And that was my problem. Super efficient at speed, super ineficient at high loads.
User avatar
Burgerman
Site Admin
 
Posts: 65051
Joined: 27 May 2008, 21:24
Location: United Kingdom

Re: UK-based robotics engineer

Postby Davross » 22 Jan 2018, 11:41

Burgerman wrote:
A modular system is what I had in mind,


Modular is compromise rather than pure design. Modular is a bad thing. Its only good for the manufacturer, or the price. Not for the design or end user.

Ask yourself why a formula 1 (no compromise ultimate solution) car is not modular.


I would argue many elements of Formula One are modular – for instance the hybrid power unit and battery packs are "off-the-shelf" and common to many of the cars. Even the engines are only made by three or four manufacturers meaning the ultimate design has to be compromised to fit the engine.

That being said, modular does create compromises and I can see from comments further down this thread that the seating system raises the CG and hence should be planned in the design from day one.

I have no preconceived ideas about what will be good, but rather coming to what I consider to be the "hive mind" on this topic.

Kev
Davross
 
Posts: 5
Joined: 21 Jan 2018, 13:37

Re: UK-based robotics engineer

Postby Burgerman » 22 Jan 2018, 13:45

I would argue many elements of Formula One are modular – for instance the hybrid power unit and battery packs are "off-the-shelf" and common to many of the cars. Even the engines are only made by three or four manufacturers meaning the ultimate design has to be compromised to fit the engine.


Because they have no choice on these specific items. Due to agreed cost reduction and rules. It results in a worse car however. One F1 engineer already said this caused many compromises. In that they cannot shape the engine casing to directly fit rear suspension points in the places they would like so they have to mount the engine and suspension seperately... Likewise he wanted the engines "dry sump" bottom cover to extend in one part under the car for rear protection and streamlining. But the rest of the car is extremely bespoke, even to the point of quite major differences for certain drivers.

So modular bad for ideal design. Always. It benefits cost, manufacturing stock of parts reduced, etc.
User avatar
Burgerman
Site Admin
 
Posts: 65051
Joined: 27 May 2008, 21:24
Location: United Kingdom

Re: UK-based robotics engineer

Postby Davross » 23 Jan 2018, 12:17

Burgerman wrote:
I would argue many elements of Formula One are modular – for instance the hybrid power unit and battery packs are "off-the-shelf" and common to many of the cars. Even the engines are only made by three or four manufacturers meaning the ultimate design has to be compromised to fit the engine.


Because they have no choice on these specific items. Due to agreed cost reduction and rules. It results in a worse car however. One F1 engineer already said this caused many compromises. In that they cannot shape the engine casing to directly fit rear suspension points in the places they would like so they have to mount the engine and suspension seperately... Likewise he wanted the engines "dry sump" bottom cover to extend in one part under the car for rear protection and streamlining. But the rest of the car is extremely bespoke, even to the point of quite major differences for certain drivers.

So modular bad for ideal design. Always. It benefits cost, manufacturing stock of parts reduced, etc.


Designs is about compromises, and in the case of F1 these are to meet the rules and budget constraints. I'm sure they impact ultimate performance, but to what extent…

In our case, we are always constrained by modules, simply because we can't design every wheelchair as a one-off from the ground up to meet our exacting specific needs. As examples of modules in your current chairs, your speed controller is modular, as are your motors, batteries, rear wheels, front casters and to some extent your chassis (albeit heavily modified). These have all compromised the ultimate design, but to a sensible/affordable degree… We don't have Ferrari's budget!

If you want to evolve this design, you have to have a basis to start from and then have feedback and improvements offered by the community. As an example, someone mentioned the use of modern materials for the chassis, which requires design work and FEM stress analysis to ensure adequate strength. Having done this, it would be daft to only fabricate one chassis – why not make the design public domain and maybe even order a dozen as a community buy to gain cost benefits.

I can immediately see a very simple area that would improve BM's current battery design. It's "module" is made up of the cells, bus bar plates and orange plastic spacers, and here is where improvement can be made. The plastic spacers are designed to create a rectangular battery pack and have been adopted for ease of use.

With 3-D printing, it would be easy to organise the cells in a different arrangement, for instance staggering each row by half a cell space, while closing the row gap. Sure the ends of the cells would no longer be flat, but the overall high would be reduced, lowering the centre of gravity. You would then need to machine new bus bar plates, but if the new spacing design was regular this is a simple CNC task.

It's not an earth-shattering change, but it's not a difficult one to make nor does it have a large cost associated – perhaps £10-20 more than using the off-the-shelf orange spacers. Isn't this the sort of improvement worth adopting?

Coming from the drone racing community, I have seen flight controllers evolve massively in just five years down to open source projects shared amongst the community. Someone initially had the smart idea of harvesting low-cost gyros from Nintendo Wii controllers and made a very low-cost flight controller from these and a domestic appliance micro. The community ran with this design and evolved it, and now you can buy a 32-bit high-speed controller board from China for less than $20 featuring current and voltage monitoring with on-board SD card data logging.

Cheers
Kev
Davross
 
Posts: 5
Joined: 21 Jan 2018, 13:37

Re: UK-based robotics engineer

Postby Burgerman » 23 Jan 2018, 15:09

If you want to evolve this design, you have to have a basis to start from and then have feedback and improvements offered by the community. As an example, someone mentioned the use of modern materials for the chassis, which requires design work and FEM stress analysis to ensure adequate strength.


Try to understand this. The basic chair is short, and very carefully configured to keep CG in the correct spot and suited to myself. Others would find it too short, and lacking stability. And too tippy. For eg it cannot use a car type seat or a heavy seat lift/tilt/power footrest setup as it would be way to top heavy and suffer instability on slopes both laterally because of soft wide tyres and on ramps or while accelerating. And the lateral lean causes caster shake issues. And its very sensitive to small changes.
So how is one chassis going to work with:
Full power seating, lift etc, and requiring the weight of lead batteries and extra length etc going to be a good chassis for a chair with far lower cg with no seating options, lithium battery etc thats 100KG lighter, and so can also be shorter? It makes no sense.
Heres another consideration. One chassis for a:
70kg chair, with a 50kg user.
A fully loaded chair with an extra 100kg of fancy seating sat high with a 100kg user? And with an extra 25kg of lead for stability in the bottom? One carefully finite analysis chair cannot do both.

Having done this, it would be daft to only fabricate one chassis – why not make the design public domain and maybe even order a dozen as a community buy to gain cost benefits.


Which scenario is this chassis supposed to be optimised or even safe for?

I can immediately see a very simple area that would improve BM's current battery design. It's "module" is made up of the cells, bus bar plates and orange plastic spacers, and here is where improvement can be made. The plastic spacers are designed to create a rectangular battery pack and have been adopted for ease of use.


I already had some, they were already available.

With 3-D printing, it would be easy to organise the cells in a different arrangement, for instance staggering each row by half a cell space, while closing the row gap. Sure the ends of the cells would no longer be flat, but the overall high would be reduced, lowering the centre of gravity. You would then need to machine new bus bar plates, but if the new spacing design was regular this is a simple CNC task.


But the actual height saving was in the order of about 15mm on a 6 cell tall pack... As it is they are already extremely compact. http://www.wheelchairdriver.com/images- ... -small.JPG The extra length just means that you cant fit the 78 cells without losing a full row from the top. So it doesent work. Or rather doesent help. Not to mention that some of my packs are a different shape designed so hat the very square shape takes best advantage!

EG download/file.php?id=4241&mode=view
For maximising Ah in a given space with 3 heights, all corners filled download/file.php?id=4452&mode=view
With not a mm wasted download/file.php?id=4460&mode=view
User avatar
Burgerman
Site Admin
 
Posts: 65051
Joined: 27 May 2008, 21:24
Location: United Kingdom

Re: UK-based robotics engineer

Postby snaker » 24 Jan 2018, 09:22

@Kev: As a powerchair user, my most concerned things are range and speed. The range is now resolved by lifepo4 conversion. Lifepo4 cells are actually even cheaper than lead batteries. With the help here plus some effort, everybody is all able to build a super big 150-200Ah@24V pack for a reasonable budget. But the speed is still a big challenge. As I know, the only way to improve the speed is to replace the current 24V controller by a 36V or 48V roboteq controller. This job is actually too complicated and requires too many deep knowledge in electronics and engineering. Only a few members here can build fast roboteq powerchairs. The rest still have to be happy with max 15kph chairs. My dream is that in future by somehow the job of building a roboteq controller would be simpler and easier. So everybody would be able to build one.
User avatar
snaker
 
Posts: 1193
Joined: 23 May 2015, 10:45
Location: Vietnam

Re: UK-based robotics engineer

Postby shirley_hkg » 24 Jan 2018, 14:03

I second snaker to put Roboteq controller into stock chairs , especially Invacare Brushless .

Higher volt means higher efficiency and torque .

Will be fine with most light weight users . cheers

Most off-the-shelf ebikes in China are 72V . A 275kg wheelchair sticking to 24V sucks . drunk2
shirley_hkg
 
Posts: 3918
Joined: 31 Dec 2010, 13:42

Re: UK-based robotics engineer

Postby Burgerman » 24 Jan 2018, 14:58

Higher volt means higher efficiency and torque .


No. Torque is limited by gearing (non on the invacare direct drive hub motors) and Amps! Amps are firmly limited by the controller as they are all current limited by software. At least anything other than hobby stuff. And they will die instead!

Worse the brushless motors pull max amps allowable even trying to do a turn in place. I measured 190 battery Amps just trying to urn left! And only 2A at full speed. Increasing volts will not increase torque at all. It will increase speed but on a brushless chair like yours speed is software limited rather than limited by the motor generator effect. Soit could go a lot faster even at 24V.

But the fact remains that they pull massive currents under stall or near stall conditions. Much worse than a geared motor. So not great for battery life/range if you are dong many ramps, hills, turning a lot as in sport etc. Because they are far less efficient under these conditions than a geared setup. Its why mine went back.

With a geared motor you can double the volts, but half the gearing to get twice the torque at the same amps. Or the same performance at half the current so much better range. As they say, Volt up, gear down!
User avatar
Burgerman
Site Admin
 
Posts: 65051
Joined: 27 May 2008, 21:24
Location: United Kingdom


Return to Introduce Yourself!

Who is online

Users browsing this forum: No registered users and 11 guests

 

  eXTReMe Tracker