Wheel width for differential drive robot

[Somchaya]

I'm thinking of building a differential drive robot, with the front 2 wheels powered by servos. I was wondering if the width of the wheels will make a big impact on the effectiveness of the turning.

Essentially, the wheels I'm considering are 6" in diameter and 1.5" in width. Will these wheels be too "fat"? I haven't calculated the final weight of the robot, but I estimate it to be around 3-4kg. Will standard servos (4.1kg/cm at 6V) provide enough torque to do the job well? (Edit: I just looked at the dynamics tutorial page, I'll do some of those calculations later, but if anyone can give a rough idea of yes/no, that'd be nice  )

Also, I'm wondering about the back wheels. I'm thinking between using a single caster wheel (maybe with an omniwheel as the caster), or maybe 2 omniwheels (so the robot looks kind of like a 4 wheeled car but with 2 omniwheels at the back) for added stability. Which would be better?

[Admin]

I was wondering if the width of the wheels will make a big impact on the effectiveness of the turning

What material are the wheels made out of? What terrain will the robot run on? How wide is your robot (wider robots have lower turning friction)? The highest friction scenario I have done is .75" width foam wheels on concrete, and didnt have any noticeable issues.

As for the caster, one is better - less friction, less to buy, less to connect, etc. Your robot will be perfectly balanced as long as mass is equal on both sides of the castor. One additional note, make sure you locate your heavy stuff on top of the castor - if not, your robot will do pop-a-wheelies when it stop-starts (I learned this the hard way ).

[Somchaya]

The wheels are made of foam, and the robot should be around 9-10" wide. I'm thinking of building a base to place my laptop on for mobot, so the terrain will be concrete with cracks I also want to possibly use the base for other things in the future too, but the terrain then will most likely be flat, or low carpet.

With such a wide base, I'm wondering if a single castor will be stable enough. The weight should be basically evenly distributed, but it may still be unstable?

[Admin]

Im gonna say it straight . . .
Ive never seen a mobot with a laptop make it more than 10ft past the starting gate

With such a wide base, I'm wondering if a single castor will be stable enough. The weight should be basically evenly distributed, but it may still be unstable?

One will work fine. Just keep your weight really low and balanced. The only way your bot can tip is if its top heavy and making sharp/fast turns.

The best way to test friction is to jerry-rig up something quick with the wheels you want, motors you plan to use, and 'simulation weights.' With just 2 or 3 hours of work you can determine if your ideas are valid or not. I did something similar my first year doing mobot - it involved lots of tape . . .

I will see you at mobot

[annoyin_kid]

what is a mobot and why do you want a laptop on a robot?

[Admin]

MOBOT is a competition at Carnegie Mellon University:
http://www.societyofrobots.com/competitions_mobot.shtml

its basically a low contrast line following competition on bumpy/hilly terrain

historically only 10% of all participants get past the first 10% of the course . . . only 2% of participants have beaten the entire course . . . so its decently hard . . .

ive done it for years . . . but this is probably my last year to compete, sigh

[annoyin_kid]

how come your last? getting too old for it?
why do you want a laptop on the mobot Somchaya?

[Somchaya]

So,the main reason is because I want to do some vision on the robot, and buying a CMUcam is pretty expensive. Coupled with the fact that I don't have a microcontroller like Basic Stamp etc, I decided to go with a laptop. And, building such a robot potentially allows me to use it for a variety of things besides mobot, so that's a big plus.

However, building a bigger robot is harder I think, and I've not really built a robot with wood/plastic from scratch before, so right now I'm working on a small scale version to get the hang of it and also point out issues I'll need to look into before building a bigger robot.

So far it's been pretty fun (read: frustrating until it works) to experiment with building and playing with servos.

[annoyin_kid]

if you want a cheap microcontroller try picaxe. its is dirt cheap and the programming software is free. the programming language is pretty easy to pick up.

[Admin]

how come your last? getting too old for it?

i plan to be traveling/living in Asia all next year
and when i get back, pittsburgh might be far from where i choose to live . . .

if you want a cheap microcontroller try picaxe

but microcontrollers suck at vision . . .

i never liked the CMUcam very much - ive built several robots with it and its only been a pain . . .
considered the AVRcam?
http://www.jrobot.net/Projects/AVRcam.html
its $100, cheaper than even the outdated CMUcam v1, but ive never used it before . . . looks better than the CMUcam . . .

[Somchaya]

Hmm.. I've never seen the AVRcam before actually, but it does look pretty decent. Still, I think it'd be nice to hook a webcam to a laptop/pc so that I can try running my own vision routines on it. Doing vision on a microcontroller just takes too long, if it's even possible.

I've actually had another friend complain about the CMUcam v1 to me too, so I guess it's generally not that good  I haven't heard about how good/bad v2 is though.. The hefty price tag tends to make me shy away from trying it.

[annoyin_kid]

how come your last? getting too old for it?

i plan to be traveling/living in Asia all next year
and when i get back, pittsburgh might be far from where i choose to live . . .

noooooooo!!!!! will you still be on the SoR forums? if not the forum would be incomplete and i built my first robot because of you 

if you want a cheap microcontroller try picaxe

but microcontrollers suck at vision . . .

they do? i never used a camera with a robot yet as i only built 1 robot. it is a budget version of an object dodging robot using a high intentisity led and ldr for the sersor. it can only dodge white walls though. 

[Admin]

noooooooo!!!!! will you still be on the SoR forums? if not the forum would be incomplete

hehe dont worry. internet cafes are like my place of relaxation when in crazy asian cities
ill still write tutorials, too

i have some connections with a robotics club in Bangkok. i visited them 2 years ago, and they had some crazy stuff. ill probably film some of it to post on SoR for everyone. same for Japan.

[annoyin_kid]

i have some connections with a robotics club in Bangkok. i visited them 2 years ago, and they had some crazy stuff. ill probably film some of it to post on SoR for everyone. same for Japan.

the man's got hookups in asia even. lol 

if you decide to drop into new zealand let me know, i know my way around the country and it is a really beautiful country

[Nyx]

I'm building a laptop based robot as well, using two webcams for stereo vision.

I think the robot hobbyist community often has a bias for microcontrollers and raw electronic hardware. The reason probably being that most of the people interested in robotics are people with a background in electronics who like to do it all by themselves.

Still, I think using a laptop is a very worthwile idea. You get alot more CPU power, alot more RAM and storage... You can run several vision routines at once without problems... The laptop has its own power source, and you can get webcams for cheap. In the end, it might even be cheaper than alot of robot kids out there. Microcontrollers, in my opinion, are just too weak for any serious visual processing. All you can do with them is very simple tricks, that don't always work.

As for most people never making it to the end of the course... I think the problem with most people is that they never test their stuff. Either that, or they test it once in a very simple setting and somehow magically expect it to always work... But honestly, there are plenty of pictures and videos out there of the mobot course... So how tough would it be to make your own test setup that resembles it and do some more serious testing.

[JonHylands]

Nyx,

I don't know if you've seen them, but there are a couple "PC-on-wheels" platform projects out there in the hobby community:

http://oap.sourceforge.net/
http://www.leafproject.org/

Personally, I share your enthusiasm about using a laptop to provide the "brains" for a robot, but I'm starting a little bit simpler than that - using a wifi link to control the "hardware" of the robot from my laptop.

You can see a lot of information about MicroRaptor, my laptop-robot project, here:

http://www.bioloid.info/tiki/tiki-index.php?page=MicroRaptor

Eventually, I hope to build a bigger biped that can carry a mini-itx based brain, but for now this will have to do...

- Jon

[Admin]

I think the robot hobbyist community often has a bias for microcontrollers and raw electronic hardware. The reason probably being that most of the people interested in robotics are people with a background in electronics who like to do it all by themselves.

im in the microcontroller camp (im a mechanical engineer) . . . i have a few thoughts on that . . .

would you trust your $1000 laptop on a prototype robot potentially crashing and flipping over occasionally?

laptops weigh a lot more than a microcontroller - this means you need bigger more expensive motors and hardware

laptops drain significantly more power than a microcontroller - bigger batteries mean more weight, again being more expensive and requiring more expensive motors. laptops win when it comes to processing capabilities, but lose when it comes to efficiency and miniaturization. depends on the application . . .

often, even when you do have a laptop for processing, you still need a microcontroller to interface with hardware. RedTeam for the DARPA Grand Challenge used a PIC16F877 to interface with a lot of their hardware. they had a humvee, with insanely fast computers, yet still needed to use a microcontroller.

you make the point of computer vision . . . i agree that a laptop would be best for that application, and a small few other processor intensive apps . . . but you dont need a 1Ghz processor for 2 servos and 2 photoresistors

what JonHylands did on his walking robot would be ideal - a microcontroller on board, but a wireless laptop doing the processor intensive stuff offboard. best of both worlds

As for most people never making it to the end of the course... I think the problem with most people is that they never test their stuff. Either that, or they test it once in a very simple setting and somehow magically expect it to always work.

and procrastination! - i see many competitors spending ~8 hours on it in the first few months, then 50 hours on it in the last week . . .
Nyx, are you from CMU too?

[Nyx]

would you trust your $1000 laptop on a prototype robot potentially crashing and flipping over occasionally?

Yes, my robot will be designed not to flip or fall (much weight at the bottom), and the laptop will have protective foam padding.

laptops weigh a lot more than a microcontroller - this means you need bigger more expensive motors and hardware

Bigger robots also mean more autonomy, potentially more mobility, a better ability to deal with uneven terrain, more strength, the possibility of extending them further (by adding a robotic arm, for example).

laptops drain significantly more power than a microcontroller - bigger batteries mean more weight, again being more expensive and requiring more expensive motors. laptops win when it comes to processing capabilities, but lose when it comes to efficiency and miniaturization. depends on the application . . .

Except laptops have their own batteries, so you don't necessarily have to power them off of the same power source as your motors.

often, even when you do have a laptop for processing, you still need a microcontroller to interface with hardware. RedTeam for the DARPA Grand Challenge used a PIC16F877 to interface with a lot of their hardware. they had a humvee, with insanely fast computers, yet still needed to use a microcontroller.

Well, you don't need a microcontroller if you have this: http://www.phidgets.com/index.php?module=pncommerce&func=itemview&KID=117062130165.92.214.11&IID=57. They also sell USB servo controllers, analog interfaces, various sensors you can hook to these kits (accelerometers, IR range finders, voltage meters, etc...).

I do think it's kind of lame, however, that there aren't more people out there who build equipment to interface computers with electronic hardware. I would like to see affordable boards with 64 digital inputs, 64 digital outputs and 16 analog inputs and 16 analog outputs.

you make the point of computer vision . . . i agree that a laptop would be best for that application, and a small few other processor intensive apps . . . but you dont need a 1Ghz processor for 2 servos and 2 photoresistors

Most likely not. My personal goal isn't to compete in a MOBOT competition, as I really see no point (plus I can't afford to go there and my robot wouldn't fit under the metal arcs). Still, I think if someone designed a robot with proper vision for that competition, he could potentially outperform alot of competitors. It's easy to think of vision algorithms that would just not fail in this type of competition.

what JonHylands did on his walking robot would be ideal - a microcontroller on board, but a wireless laptop doing the processor intensive stuff offboard. best of both worlds

My problem with that is the lag, the bandwidth limitations (important when doing vision processing), and the limited range of wifi (your robot just won't go very far from you and your laptop). Obviously though, my ideal robot for something like that would be something like a modified RC car, and it would be difficult to fit a laptop on that.

and procrastination! - i see many competitors spending ~8 hours on it in the first few months, then 50 hours on it in the last week . . .
Nyx, are you from CMU too?

No, I just know how people are. Well, you might not like my attitude, but I personally don't see much point in going to a competition like that if you're not making efforts in the direction of winning it (what's the point of participating if you're not trying?).

[Admin]

my robot will be designed not to flip or fall (much weight at the bottom)

expensive robot accidents do happen
for example:
http://www.defensetech.org/archives/images/drone_flip.JPG
http://www.redteamracing.org/index.cfm?method=gallery.viewGallery&page=8

Except laptops have their own batteries, so you don't necessarily have to power them off of the same power source as your motors.

but thats still an extra battery, and extra weight . . .

Well, you don't need a microcontroller if you have this: http://www.phidgets.com/index.php?module=pncommerce&func=itemview&KID=117062130165.92.214.11&IID=57. They also sell USB servo controllers, analog interfaces, various sensors you can hook to these kits (accelerometers, IR range finders, voltage meters, etc...).

those are all microcontrollers, just much less functional ones since you cant reprogram them . . .

Still, I think if someone designed a robot with proper vision for that competition, he could potentially outperform alot of competitors.

the guy that currently holds the time record used a 90's handheld iPaq with an old webcam
the second highest record is a guy that used a microcontroller with the CMUcam (he also invented the CMUcam)

Well, you might not like my attitude, but I personally don't see much point in going to a competition like that if you're not making efforts in the direction of winning it

Well, I do it for winning and for fun. If I couldnt win, I'd still do it for fun - plus its educational. I think most students start early, but then get bogged down with day-to-day classwork and exams. Priorities change, and dedication/time-management are skills they are still learning . . .

[Nyx]

expensive robot accidents do happen
for example:
http://www.defensetech.org/archives/images/drone_flip.JPG
http://www.redteamracing.org/index.cfm?method=gallery.viewGallery&page=8

Well, my robot has nothing to do with car races

but thats still an extra battery, and extra weight . . .

Well, if you can carry a laptop, chances are, you can carry a laptop with its battery. I've seen some pretty cheap ($100) 1:10 scale RC hummer toys. You could probably fit a laptop and one of those fidget cards inside of them pretty easily, and they would have no difficulty carrying it.

the guy that currently holds the time record used a 90's handheld iPaq with an old webcam
the second highest record is a guy that used a microcontroller with the CMUcam (he also invented the CMUcam)

An iPaq is pretty much a very small PC. So you're confirming my point A pocket PC would be a good second choice to me, except that they can be pretty limited in terms of connectivity/development. And still, you get at most a quarter of the CPU power and RAM of a laptop. My ideal choice, given the money, would be a sony VAIO with an 11.1" screen. Those are very small, light, and have a very good battery autonomy (sometimes as much as 6 hours). But if you're on a tight budget, some people sell laptops with broken screen on ebay for like $100. You can get an old 500 MHz+ with 128 MB RAM for cheap, install linux and a WiFi card on it, and start programming.

Well, I do it for winning and for fun. If I couldnt win, I'd still do it for fun - plus its educational. I think most students start early, but then get bogged down with day-to-day classwork and exams. Priorities change, and dedication/time-management are skills they are still learning . . .

Well, I agree about classwork. This is why I plan to build my robot in the summer. Right now I'm mostly planning and buying parts, doing some reading.

[JonHylands]

My problem with that is the lag, the bandwidth limitations (important when doing vision processing), and the limited range of wifi (your robot just won't go very far from you and your laptop). Obviously though, my ideal robot for something like that would be something like a modified RC car, and it would be difficult to fit a laptop on that.

I don't imagine there is going to be a lot of lag most of the time. I won't be sending video over the wifi link - the cameras are 2.4 GHz wireless CCD cameras, and I will have two USB capture cards on my laptop feeding into a separate application (most likely RoboRealm to start with) that will do the preliminary vision processing.

I know my range will be limited, and I'm okay with that - what I'm trying to build is the framework and AI system upon which my real Raptor robot will be built, and it will have a full-blown on-board CPU.

- Jon

[Admin]

Nyx, your argument suggests a PC is better than a laptop

The reason why I perfer to put a microcontroller on a robot instead of a laptop is the same reason you perfer a laptop over a full fledge PC.

In the end, what is optimal all depends on what the robot is required to do . . . no?

[Nyx]

Nyx, your argument suggests a PC is better than a laptop

Laptops are PCs

The reason why I perfer to put a microcontroller on a robot instead of a laptop is the same reason you perfer a laptop over a full fledge PC.

I actually initially intended to use a regular ATX motherboard with a WiFi card. It would have taken up a bit more space, but for a big robot, it doesn't really matter. I could easily have fitted a hard drive and a power supply in there. The reason I'm using a laptop is that it's a bit easier and I already have a pretty powerful one.

In the end, what is optimal all depends on what the robot is required to do . . . no?

Of course. My main gripe is that you seem to be directing all hobbyists towards microcontrollers. It also seems that all robot kits are made for microcontrollers. Almost nobody makes kits that are designed to use laptops and/or PDAs.

I think that the fact that 99.9% of robotics hobbyists use microcontrollers is essentially limiting the amateur research perspectives in the field. I think more could be achieved by hobbists if they used more powerful computers, better development tools, giving them a chance to focus more on the algorithmic side of robotics, instead of only the electronics. Of course, you could claim people in universities do that, but from what I've seen of academic robotics labs...

My robots is basically an experiment, a research tool of my own. I can't say that I will be able to compete with academic research labs, but I'm fairly confident I can make it do interesting things. Things that just can't be done with a microcontroller.

[JonHylands]

Nyx,

What you're describing is the entire push of MRS (Microsoft Robotics Studio).

They are pushing the same metaphor, for the same reasons.

Evolution Robotics is another company that does the "mobile-laptop" route, but they have priced their software way out of reach of the hobbiest market.

- Jon

[hgordon]

Of course. My main gripe is that you seem to be directing all hobbyists towards microcontrollers. It also seems that all robot kits are made for microcontrollers. Almost nobody makes kits that are designed to use laptops and/or PDAs.

A laptop is not a natural fit for this type of application -- if you want lots of i/o, you need a computer with a real expansion bus, such as PCI or PC104.  There's no serious market for using laptops in industrial control, as the devices are neither sufficiently robust nor expandable.  On the other hand, there are a wealth of expansion options once you start looking in the industrial computing direction, and you gain the benefit of powerful processors and good development tools.

[Admin]

The majority of robots at CMU use laptops . . . but I think thats because 90% of the people in the Robotics Institute are computer science majors and would rather stick to what they know . . . much of it is understandable, as CMU does heavy research on vision, SLAM, and laser rangefinders . . . and they dont plan to market their research so it doesnt need to be cost effective . . .

[Somchaya]

Well, I'm a CS major and I intend to use a laptop for a robot for many reasons. One of the main reasons, as said above, is that laptops are generally easier to code in compared to learning to use a microcontroller.

Also, a big factor is that I've been searching around for an easy way to interface a camera/webcam to a robot, and besides putting them on PDAs and such, a laptop is the next easiest. I considered using a PDA, but the cost of the camera, the interface and all that doesn't help. A laptop can easily interface with a camera, and through things like servo controllers, it can interface with servos and other sensors as well!

So, while bulk and weight is definitely a factor, using a laptop saves development time (getting used to programming a microcontroller), some money getting the parts I need (since I don't own a microcontroller), and provides lots of memory/CPU. The only trouble is that it's dangerous to let my laptop run around on a potentially self-destructive robot, but hehe, I'll keep close watch and stay really close to it in case it crashes, and hope that it'll do.

[trigger]

Well, I'm a CS major and I intend to use a laptop for a robot for many reasons. One of the main reasons, as said above, is that laptops are generally easier to code in compared to learning to use a microcontroller.

What about C on an AVR? Using gcc? It's not a big jump ...

Also, a big factor is that I've been searching around for an easy way to interface a camera/webcam to a robot, and besides putting them on PDAs and such, a laptop is the next easiest. I considered using a PDA, but the cost of the camera, the interface and all that doesn't help. A laptop can easily interface with a camera, and through things like servo controllers, it can interface with servos and other sensors as well!

I've never interfaced an mcu with a camera, so I'm not sure about this one. But it can't be that hard. There are well-documented cameras out there for mcus.

So, while bulk and weight is definitely a factor,

More expensive servos/motors, batteries, materials, etc... a major factor!

using a laptop saves development time (getting used to programming a microcontroller), some money getting the parts I need (since I don't own a microcontroller), and provides lots of memory/CPU. The only trouble is that it's dangerous to let my laptop run around on a potentially self-destructive robot, but hehe, I'll keep close watch and stay really close to it in case it crashes, and hope that it'll do.

to each his own, I suppose.

[hgordon]

I've never interfaced an mcu with a camera, so I'm not sure about this one. But it can't be that hard. There are well-documented cameras out there for mcus.

Somchaya is right - it is a LOT easier to interface a camera to a PC / laptop than a microcontroller, as microcontrollers rarely have sufficient memory for full frame storage, building a bus interface is complicated, USB Master mode is generally not an option (so you can't use USB cams), serial interfaces are slow, etc.  CMU-Cam / AVR are specialized devices with low resolution, and do not provide images which are particularly useful for serious image processing.

I understand the appeal of using a laptop, and hope you don't suffer any damage in the process.  I personally prefer to develop with microcontrollers, but I'm using a 32-bit processor with adequate memory and GNUARM development tools.

[Admin]

found this today . . .

kinda reminds me of the night before the MOBOT competition . . .

http://www.doonesbury.com/strip/dailydose/index.html?uc_full_date=20070129