PINNED - Roboteq Controller - developing for powerchairs

[LROBBINS]

Very good. I think. But I don't understand it

Uh oh, or as someone else once said "Houston we have a problem". Before I can fix it, if I'm indeed able to, I need to know what part of this you don't understand.

Do you not understand the purpose of this Programmer node?

You understand its purpose, but don't understand the list of User Settings?

You understand both of the above, but don't understand my attempt to illustrate how to use it?

Ciao,
Lenny

[Burgerman]

No mostly I cant read it!

Jpeg compression artefacts

If you save these as gif files they are tiny, completely clear and no distortion but fewer colours.

[LROBBINS]

OK. Instead of posting separate figures, I had made a pdf of the whole thing and pdf's compressions really, really sucks. I'll e-mail you the pdf, and put together something more readable for here.
Ciao,
Lenny

[LROBBINS]

I'll start posting again, pasting text and uploading figures. Will have to be several posts.

Using the Programmer node to set up the WheelchairCAN system and adjust User Settings

**************************************************************************************************************
The Programmer node is an adapter that connects between a USB port on a PC and the CAN bus. It is used to do the initial setup of the WheelchairCAN system, to calibrate or re-calibrate the Joystick and SpeedPot and to set a wide variety of User Settings
**************************************************************************************************************
Connect the Programmer node to the PC and to any unused CANbus socket.
Open the Programmer sketch in the Arduino environment. WheelchairCAN user settings are in the UserSettings.h tab:

User Settings 1.jpg (226.65 KiB) Viewed 6030 times

User Settings 2+3.jpg (227.56 KiB) Viewed 6030 times

Make any changes to the User Settings that you want and download this from the PC to the Programmer. If you save the file with a new name it will be put in a new folder without deleting the original. If you do this each time you make changes, you will have a complete record of what you've done.

Note that some Roboteq-specific settings (motor resistance, acceleration parameters) are not here. There is still some refinement of the acceleration and/or motor compensation algorithms needed in the Roboteq script, but once John's done some more testing (when and if England's weather improves) they'll be moved here as well.

Now, open the Serial Monitor, and you will see the following information:

Prog start screen.jpg (84.81 KiB) Viewed 6030 times

Before Sending a character to start the programmer, turn ON the WheelchairCAN system if it's not already on. If you forget, and send a keystroke before Programmer node is plugged in, or while WheelchairCAN is asleep, or if WheelchairCAN was OFF, don't worry. You will get this screen that will tell you what to do:

[LROBBINS]

Prog asleep screen 2.jpg (111.15 KiB) Viewed 6029 times

Programmer will now detect if this is the first time that this WheelchairCAN system has been programmed, or if the Joystick and SpeedPot have never been calibrated. If it's a new system, it will automatically send all your User Settings to WheelchairCAN and then tell you how to calibrate the Joystick and SpeedPot. (Until the first programming is done, Master has some provisional values that work well enough that the system won't "crash".)

If this is not the first time programming the system, you will start seeing a series of questions, the first two asking whether you want to re-calibrate the Joystick and SpeedPot or adjust the voltage sensing to match your actual battery voltage. Just follow the instructions.

Whenever you see Y/N it means enter 'Y' (or 'y') for "Yes, I want to do this" or any key except 'Y' or 'y' if you want to skip this step. Volt meter calibration is an exception since you have to enter a number to adjust it. You may have to try several values of the Calibration factor to get WheelchairCAN to match battery voltage, so these three lines will repeat until you enter any letter to signal that you're done.

Prog screen.jpg (227.31 KiB) Viewed 6029 times

In the above, I've answered N to all of the questions, and WheelchairCAN already had the same User Settings as in the Programmer's UserSettings tab so the output is quite short. Each month, a new log file will automatically be created. You can look at the latest one, or at any of the earlier ones, and if you chose to look at an earlier one you will get instructions about how to tell Programmer which one you want.

Even if this is the very first time you've turned on WheelchairCAN there will be a log file showing the date and time of startup, the amount of time elapsed since the chair was turned on, and the total amount of time that the system has been ON. Time spent sleeping or turned off doesn't count. The log files are stored in a micro SD card in the Display node. If you want to have a copy on your PC, you can copy what is shown in Serial Monitor and paste it to a text editing program - the Arduino IDE does not have a way to save the serial output. (You could also pull the SD card, but for system reliability that's probably not something to do routinely. It also has all of the display graphics and error logging messages stored on it!).

Whagt follows are a couple examples of other Programmer sessions.

I. Programming WheelchairCAN for the first time:

NewMCU all.gif (171.42 KiB) Viewed 6029 times

Notice that the current log file shows initial startup date and time with 0 RunTime and TotalTime because this is a "fresh" WheelchairCAN. There's also a Read amps error message because the Roboteq is sometimes busy doing other things and doesn't "see" the battery amps CAN query. Not to worry, in 3 seconds Master will ask again and will update the Amp Hour bar graph on Display. There's a "dummy" log file for 12/2014 on the SD card just to illustrate how to retrieve an older log file.

[LROBBINS]

II. Programming new user settings:
Here I've changed a couple user settings and downloaded the program to the Programmer node:

UserSettings changed.jpg (88.46 KiB) Viewed 6028 times

and here's what I saw when I opened Serial Monitor

Change User Settings 1.jpg (191.2 KiB) Viewed 6028 times

Oops. The above is an old figure - I've since managed to simplify getting things restarted if one forgets to turn on WheelchairCAN before sending a keystroke. See an earlier image for the correct text.
Notice that WheelchairCAN was asleep and I forgot to wake it before entering a keystroke. I awakened it by pressing the ONOFF switch and the Programmer procedure started.I mostly answered No to each of the questions, but then Programmer detected that two user settings had been changed. It showed me each of the old and new values and asked me whether I wanted to download each change to WheelchairCAN. I said 'y' for one and 'n' for the other.

Change User Settings 2.jpg (193.11 KiB) Viewed 6028 times

[LROBBINS]

Here's what the log file shows:

restart.jpg (99.53 KiB) Viewed 6027 times

WheelchairCAN was awakened two hours after the first Startup and 20 minutes later there's was a Left brake open error logged - the display also showed a Brake fault message. After shutting off WheelchairCAN a while later and then turning it back on, RunTime was once again 0, but it now shows 33 minutes total time since the system was first activated.


WHAT TO DO IF ALL ELSE FAILS

Although I have tried to prevent errors from blocking the system, I have probably not thought of every possible failure mode. If WheelchairCAN or the Programmer become completely unresponsive, you can force a reset by doing the following:

1 - unplug Programmer from the CAN bus
2 - open the Emergency Stop switch or plug
3 - reconnect the Emergency Stop
4 - restart WheelchairCAN using the regular ON/OFF push button on master
5 - plug in Programmer and proceed normally


P.S. Working in the warmth of my bedroom, I am running this from a 19V computer brick. Measured voltage is 18.7V (or 187 deciVolts as calibrated here). Down in my cold shop where I have a couple old motors, I run it off a couple AGM batteries.

[Burgerman]

Just been looking at the emailed PDF. Looking good and like a lot of work!

As for the bit about my testing and some data adding, you will be pleased to know its -5C everything is ice and a light dust if frozen snow...



So I am setting up my new Revolectrix PL8v2 (two off) charging station next to my PC.

[LROBBINS]

We are at about +5 now, and were at -5 overnight, but no snow.

BTW, I of course can't ever leave well enough alone and have already made another change in the Programmer node. I didn't like that if one forgot to turn the Wheelchair on before starting the programming, you had to unplug Programmer, start the chair and then reconnect the cable. Too a couple days and lots of head scratching, but now the message is:

Startup is taking too long:
Is the Programmer plugged in to a CAN socket?"
Is WheelchairCAN asleep?"
Did you forget to turn WheelchairCAN on before sending a keystroke?"
----- Just plug it in, wake it up or turn it on and programming will continue -----"

Ciao,
Lenny

[Burgerman]

These things are never completely finished. I spent 3 years with my programmer friend doing C programming in DOS when finishing my rolling road software. And it was still not done!

[Burgerman]

http://www.wheelchairdriver.com/BM3-con ... RC-BM3.mp4

The acc issue.
You have seen this before. Turn up sound, and listen to the acceleration as the motors get moving. At 1 min 25 secs.
It accelerates just as hard (if not more so) right before I give it back stick. This is about 10 or 11 mph. This is acc set quite low. It accelerates the first few inches to a foot quite slowly. Wiggling the stick about does nothing. But it reaches the top speeds you see here by just around 2/3rds pulsewidth.

So it takes a while after you gun it for the pulsewidth to leave the deadband, and then start to develop enough current to move the chair. Once it does, then it builds fast. Seems like someone hits the Nitrous button! The actual stick can be wiggled about to full power and back to zero 2 or 3 times a second if you are stationary, and the chair doesn't move. Push it and wait and whoosh. Like a two stroke with tall gearing hitting its power band.

If it is accelerating this hard while you are in it, flips you out. And hitting reverse isn't fast enough to stop the acceleration. That takes maybe .3 of a sec to respond. before the pulsewidth reduces enough to drop the front. About .2 sec too late!

It does this because a 6mph chair set to the same acc rate, needs 100 percent pulse width to hit its 6mph. This one is already going 6mph by 35 percent.

So if I want the chair to reach 6mph in the same time as a stock chair I need the pulsewidth to reach 35% in the same time a stock chair controller reaches 100%. Its this and the fact that I can still get full 150A torque at that speed, that means I have to set acceleration 3x slower. This means its very sluggish at the instant "jump" off the line. The bit where it should lift the wheels over a curb for e.g.. This is where my delay comes from. Its because I must set it low. Because a 50% pulse-width is already 8.5mph (or more due to motor compensation help)!

Its the daft rate of acc at high speed that is the problem. It goes like my overpowered drag bike. (Same acceleration at 170 mph as at 70 mph limited by tyre only...) Potential 14kw? Doesn't sound much but It stops me turning up acceleration to get rid of the delay at 0 to 2 mph from a standstill. It either one or the other. It really needs a way of reducing Current as the chair speeds up - as a stock 24v chair does.

Trust me its nowhere near as bad as I am making it sound - that's just the easiest way to explain it. Most people would be just amazed at how much better it is than a stock powerchair. But if anything went wrong with control this would break your legs or worse!

[LROBBINS]

John,

I doubt that this will be a complete solution, but I think that there's something you can try that's already in the existing script. I'm basing what I'm saying here on the following mental picture: At very low speeds, current draw (and hence Motor Compensation) will be high. As speed increases, current draw will fall off, at least until you're getting near maximum possible motor output.

There's already a parameter in the script that can be set to increase acceleration when Motor Compensation (current) is high, and let it go back down as current flow declines. At least until you get to the point that current is again rising, it should let you have higher acceleration at the beginning and gentler acceleration as speed increases.

Now, I have so many versions of Roboteq scripts here that I can't keep track of them, so I'd like you to search in yours for two lines:

(1) AccelMotorComp = 30 (or some other number)

and

(2) BaseM1Accel = M1Accel + AccelMotorComp*abs(M1Comp)/10

Does your script have (1)?

In (2) is the sign + or a minus?

If your answers are YES and PLUS, here's what I suggest you try.

1 - Reduce M1Accel and M2Accel to as low a rate that still lets the chair move at all. Is the acceleration at speed more reasonable now?

2 - If acceleration at speed is now too slow, increase M1Accel and M2Accel by little bits until you get the acceleration AT SPEED that you want. It will still suck at low speed.

3 - If that tames things adequately at speed, now increase the value of AccelMotorComp. This will boost acceleration rate whenever current draw is high so will be especially effective at low speed.

4 - If it is now better at low speed, but still accelerating too hard at high speed, repeat 1-2-3 as many times as needed to get the balance you want. Always change M1Accel and M2Accel first based on performance at speed then increase AccelMotorComp to get more oomph at low speeds.

Does this help? I think it should, though I'm nervous about what might happen when going near full out and current again starts to rise, but I think this would be a decent first step for figuring out what needs to be done.

(There is also DecelMotorComp, but let's work on one thing at a time - you may still want strong decel at high speeds so you can stop quickly.)

Ciao,
Lenny

[Burgerman]

Current script here:

http://www.wheelchairdriver.com/BM3-con ... Script.txt


Will take a look as soon as its warm, dry, and I can go onto a big flat field with my laptop... Need to spend a couple of hours testing where I can fine tune everything and feel the affects. And also test the above.

Deceleration from high speed - at least if you do this directly as you are accelerating - seems to "come in" slowly, then build. Strongest deceleration happens as you are stopping. It feels as if you are moving the stick from full forwards to centre slowly. The opposite to how accelerating feels. I may be over critical... But really need to spend some time fine tuning things.

I have to add that if I am racing my friend on an electric bicycle It massively out performs him in acceleration, deceleration and turning tight. Its a mass of tyre screeching wheel spinning and hooliganistics when we were testing in a huge college car park last October. Which is marvellous! Like driving a mini racing car! It even sounds good!

[LROBBINS]

OK, AccelMotorComp is in your script (and set at 0) and the sign where it is used is +. From the way you describe deceleration from speed, it really sounds the same as what's happening during acceleration. Decelerating from high speed, current flow will be highest at the beginning and get less and less as the chair slows, and deceleration is not strong enough when current is high, and may be too little when current is low. First let's see if we can get this sorted out for acceleration and we'll then know how to proceed with deceleration as well.

So, reduce M1Accel and M2Accel to where you can barely move from a standstill and try increasing AccelMotorComp. I'd try 20. With no load, I could see & hear an effect of AccelMotorComp=20 and the effect should be stronger with load.

Given that in the film the chair was under RC control, and the acceleration ramping was pretty obvious, you should be able to try adjusting the script the same way - without your valuable butt sitting in a potential rocket.

Ciao,
Lenny

[Burgerman]

The problem with that is that if it runs away I cant pull the plug...

I will wait till I can test while sat in it somewhere safe and open, and soft!

[BounderGimp]

I tried to play your mp4 file, computer says it's corrupt and won't play it, perhaps just my Computer in a bad mood?

[Burgerman]

Well I just tested it from my phone, my EX GF computer, my desktop, and my laptop! Working here.

Maybe you need to download the correct codecs? What PC?

Download to desktop before trying to play it.

[BounderGimp]

Thanks, I downloaded the file and it plays fine, don't know why I got the corrupt file message, I might have removed some software or code by mistake, found this information on removing microsoft controls from the computer, "trusted installer" so called, I think I got a little carried away!

http://helpdeskgeek.com/windows-7/windo ... installer/

Below is my computer.

http://h10025.www1.hp.com/ewfrf/wc/docu ... 872428#N67

[ex-Gooserider]

Would one of these battery meters be useful on a stock (lead brick power) chair as a possibly more useful indicator of charge state than the pretty useless row of LED's?

Also, is this CANbus programmer that Lenny is describing good for any Roboteq chair, or only one that has Lenny's full Arduino stack? It wasn't clear from the description, but at least some of what he was saying implied the presence of the other nodes...

ex-Gooserider

[LROBBINS]

The programmer module is for adjusting user setting, calibration, and log reading, so yes, it works with the Master, Aux and Display modules (the Display module contains the SD card) and the CAN-scripted Roboteq of the WheelchairCANbus -- it is not a replacement for Roborun. Ciao, Lenny

[Williamclark77]

I don't know that anyone else would be interested in this, but I added a second linear pot to adjust the accel rate. I had written code to adjust the accel based on the speedpot setting. That works fine. However, if I want top speed, for example, going up steep hills or general use, without a fast enough accel to wheelie and be so snappy, I can't. With the accel rate fairly low, my chair is as smooth as any stock powerchair. But then it doesn't wheelie easy enough. Fast enough rate for easy wheelies and she's a bit too jumpy for indoor or steep hills.

I picked up this linear potentiometer. Crap cellphone picture. I still have to actually install it into my already overcrowded joystick housing, but it worked on the jack testing.



I added these lines of code in various places

Code: Select all

DIM AccelPotInstalled AS Boolean


Code: Select all

AccelPotPin = 5 'analog input pin used for AccelPot


Code: Select all

AccelMinimum =400 '0-1000 Percentage of M1Accel and M2Accel that accel rate will be reduced by.


Code: Select all

AccelPotInstalled = TRUE  'Is accel  potentiometer installed?


Code: Select all

AccelMinimum = AccelMinimum


And then I modified Lenny's accel mixing formula to be

Code: Select all

   IF (AccelPotInstalled=FALSE) THEN
   BaseM1Accel = M1Accel + AccelMotorComp*abs(M1Comp)/10
   BaseM2Accel = M2Accel + AccelMotorComp*abs(M2Comp)/10
   BaseM1Decel = M1Decel + DecelMotorComp*abs(M1Comp)/10
   BaseM2Decel = M2Decel + DecelMotorComp*abs(M2Comp)/10
 ELSE
   AccelPot = GetValue (_AIC, AccelPotPin)
      IF AccelPot < AccelMinimum THEN
      AccelPot = AccelMinimum
   END IF
   BaseM1Accel = (M1Accel*AccelPot)/1000 + AccelMotorComp*abs(M1Comp)/10
   BaseM2Accel = (M2Accel*AccelPot)/1000 + AccelMotorComp*abs(M2Comp)/10
   BaseM1Decel = M1Decel + DecelMotorComp*abs(M1Comp)/10
   BaseM2Decel = M2Decel + DecelMotorComp*abs(M2Comp)/10


What this does is modify the accel rate by percentage. 0 to 1000. 400 = 40% of your max accel, 600 = 60%, etc. I could probably add a line of code to make 40 40%, 60 = 60%, etc instead of the 3 digit scale.

I must say I'm still tickled pink with how awesome the chair is. Miles upon miles of pounding it and not one broken part, gotten stuck, hardware failure, nothing.

Quick video I recorded the other day with my new Sony Action Cam. Much better than $$$ worth of dSLR hanging on the side of the chair. The fisheye distortion does make the caster fork look horribly bent though!

http://www.willsjunk.com/Other/WillChai ... -JqKT3Hr/A

[Burgerman]

So you can change accel on the fly? Cool.

What settings are you actually using for acc / dec etc?

Can I look at a copy of your current script?

To Burgerman@ntlworld.com if its easier. Thanks. I am looking to compare.

[rustyjames]

Will, that chair works fantastic! Looks like fun!

[LROBBINS]

Will,

Would you try the following and let me know what happens?

(1) Set your accel pot where you find things comfortable for basic driving.

(2) Increase the values of AccelMotorComp and (start with changes of 10 at a time, but you can go up as high as comfortable even over 100).

(3) if initial deceleration from high speed is sluggish, do the same for DecelMotorComp.

Doing (2) should boost acceleration at high current such as when trying to wheelie (at least I think it should do this).

Ciao,
Lenny

[Burgerman]

>> Doing (2) should boost acceleration at high current such as when trying to wheelie (at least I think it should do this).

What's the danger of this plus compensation causing a rather wild runaway flip?

[LROBBINS]

If you don't overdo it, I don't think it will run away.

As soon as you accelerate a bit current draw goes down and the boosted acceleration rate goes down. You will be headed toward a higher final value than otherwise because compensation raises the instantaneous command value, but that's the effect of compensation alone and once current starts to drop you won't be accelerating toward that value any faster than you would have been without the boost of accel. At least I think that it will behave this way.

As the only motors I have here are unloaded I have no way to test this. Next week I'm going to add some connectors to Rachi's chair so that I can swap between the DX system and my CAN+Roboteq system, so if that goes well I should be able to do some real testing too. But I have to do this in a way that I can avoid taking her chair out of commission except when she's napping.

Ciao,
Lenny

P.S. I wouldn't mind having an old clunker chair for testing even though I don't know where I'd put it, but the chance of finding one in Italy is about nil. Powered wheelchairs are few and far between here - I know of 3 in the city of Siena and nearby suburbs, about 70,000 population total, and one more in a town further south.

[Williamclark77]

BM, yes it adjusts accel rate on the fly. Set your M1 and M2Accel to whatever you want the max rate at and the AccelMinimum to the slowest you want. For example, M1 and M2Accel on 5000 (500 rpm/s) and AccelMinimum at 500 (50%) gives 500 rpm/s acceleration at pot full forward. Pot all the way back gives 250 rpm/s. Pot in the middle gives 375 rpm/s of acceleration.

Lenny, I will try that tonight or tomorrow. I've already gotten back out of the chair and started pulling my joystick pod apart to install the pot. Hopefully Sommer won't slap me when I ask her to help solder inside that housing again here in a little bit. She did get a big bouquet of purple orchids delivered to her work from me yesterday. So, I SHOULD be good to hit her up for it.

I THINK doing as you suggest will work fine for general use and be better in most situations, especially if you don't want the hassle of adding an extra pot. I'll have to see on hills. I honestly haven't tried those settings much before.

Rusty, thanks! I've beat this chair like a rented mule. I pushed it hard today. Hard enough that I started to smell toast. The motors were so hot that you couldn't hold your hand on them long. Lots of low speed heavy loading generates some heat! I recorded video. Haven't watched it yet to see if it's interesting looking enough to post.

[Williamclark77]

Will,

Would you try the following and let me know what happens?

(1) Set your accel pot where you find things comfortable for basic driving.

(2) Increase the values of AccelMotorComp and (start with changes of 10 at a time, but you can go up as high as comfortable even over 100).

(3) if initial deceleration from high speed is sluggish, do the same for DecelMotorComp.

Doing (2) should boost acceleration at high current such as when trying to wheelie (at least I think it should do this).

Ciao,
Lenny

Lenny, AccelMotorComp does exactly as you think. Adjusted up to 70 allows me to do a quick wheelie over stuff, even with the base accel rate low enough to be very smooth. As I said above, this setting you wrote in would be better for most users and general use, especially to avoid the "fun" of wiring in a second pot.

However, I'm not sure if you watched any of my videos, but I use this chair on very steep hills. Hills really too steep for being on in a chair. I need to be able to accelerate slowly and build speed without going over backwards. Hence why I added this pot to manually lock the accel rate low. AccelMotorComp works perfectly, which is the problem. My addition of slowing the accel based on the speedpot position also works, but I want the manual control so I can build full speed on the approach without too much ooomph if I slow midway up.

My Permobile actually feel safer on these same hills due to it's further forward center of gravity and heavier base - as long as I go up at angles instead of straight ahead to lessen the actual incline. It can only get up one hill before it overheats and flattens the batteries to the point I can't get back to my car though.

But yes, AccelMotorComp works exactly as you expected.

[LROBBINS]

I need to be able to accelerate slowly and build speed without going over backwards. Hence why I added this pot

Got it. Maybe John would like to have this for his pub ramp too.

AccelMotorComp does exactly as you think. Adjusted up to 70 allows me to do a quick wheelie over stuff, even with the base accel rate low enough to be very smooth.

Good. That's what I was hoping it would do, and what I hope will give John quick starting without the super zoom to full speed. I think that DecelMotorComp will also prove useful - otherwise with a reasonable, non neck-snapping, decel rate it might take too long for initial deceleration from high speeds.

Ciao,
Lenny

[Burgerman]

I hope so too. If it ever warms up, stops freezing or raining and stop pouring corrosive crap on the streets, we will find out! So not for a few weeks.