I’m looking at getting a Tri-Track robot along with a Sabertooth 2X5.
The aim is to make a fully autonomous robot which just wanders around.
I may get a arm or something later to give it a real purpose but this should keep me entertained for awhile.
Control will be done by a ATmega128 as the brains and a ATmega32 to process all the sensor data.
Motor control will either be by PWM or serial. Not sure yet which one is better.
Sensors (at the moment) will be 5 ultrasonic range sensors (two angled pointing forwards, one for each side) and a infrared range sensor pointing straight forwards.
It will be powered by one or two 12v sealed lead acid batteries.
I’m a software guy hence the beefy controllers and sensors.
Some questions: ] Is my plan at the moment good and are there any possible pitfalls with it?/:m] ] Will both motors drive at the same speed with the Sabertooth or will I need to take that in to account in software?/:m] ] Whats better: PWM or Simple Serial? The ATmega128 has dual serial which allows one for debugging and using PWM with it is also simple./:m] ] What other ways are there to get the same effect as the ‘Quadrature Motor Encoder’? I’m in Australia and its not easy to get parts. I’m actually getting the gear off eBay (new) as the Australian distributors are kinda slack. I’m thinking a photo-interrupter may do the trick./:m]
Thats all my questions so far. I’ll post more if I think of them.
Although the chassis uses “universal” motors, meaning they are designed to rotate at the same speed in both directions, they are always a wee bit faster in one direction. So you will need to account for this in software. Hope this helps.
if you have the sabertooth that accepts serial data, and you have a spare UART on your processor then it may easier to just send serial commands rather than generate the PWM signals (not that this is hard, just maybe more complicated than sending strings out the uart.) The serial may give you some other status type feedback info options that you will not get with the PWM output as those are open loop.
friction in the gearboxes and with the tracks will be different so no you will not get exactly the same output from both tracks. again this is open loop and to get closed loop response you need an encoder of some sort. It doesn’t necessarily need to be a quadrature encoder though. If the motors have the shaft for the encoder you can use a simple half disk and photo interrupter to count shaft revolutions. Since you know whether the motor is commanded forward or reverse the count should be reasonably accurate for determining position but there may be small start and stop errors.
I personally would use a Ping ultransonic pointing straight forward instead of the IR. Just curious, but why did you spec it that way? In my experience the Ping outperforms IR in every situation.
PWM and Serial are both fairly simple to use with the ATmega128.
If I’m sure that I wont need the second USART then I will go for serial.
I guessed the motors wouldnt drive the same amount.
I dont have a lot of equipment/experience at manufacturing parts so I’d prefer to spend money on a professional or dead simple solution.
Making my own encoder would be last resort.
If ultrasonic sensors are better then I’ll go with 6 and figure out a way to separate the front ones (probably by firing them separately).
I just assumed IR was better. It also has slightly further range.
you may be able to buy the encoders directly from usdigital.com but it will probably be more expensive than through lynxmotion or their distributor network. you can certainly send their sales e-mail a question about how much it would cost. customs… harder to know what to expect.
Really? I ran a series of tests and found that not really to be the case. The Ping detected everything the IR detected and reported more accurate and consistent readings. The IR could not detect smaller or narrow objects and the readings were all over the place. YMMV!
In the future I want to conduct tests that are more scientific.
The PING is certainly easier to work with. I prefer the PING to IR rangers, and the PING does not require analog inputs, which I could use for my accelerometer.
I think the PING gives quicker response and more accurate results. It also doesn’t require a lookup table like the Sharp IR Rangers need. I am going to replace all the Sharp IR Rangers on W.A.L.T.E.R. and see how he runs with that setup.
Have you looked at the LV-MaxSonar - EZ1 Sonar Range Finder? I have four of them and they work equally as well if not better than the Ping. You can find them at www.maxbotix.com.
Actually where I was going with that was that you guys appear to buy the encoders in bulk and have less of a markup than us digital does as the encoder with the cable cost less from you than if you buy the same thing off of the us digital website.
I forgot totally about the incredible ups handling fees you get socked with when shipping internationally. the canadian guys know all about that and I’m sure its even more ridiculous going to au.
I found the next best thing to a Aussie supplier: A Canadian supplier with cheap shipping.
$30 for 2 week shipping for 6 EZ0’s, two encoders and a few other goodies.
I’m overseas for Christmas holidays (2 months - yay) and I’m planning everything so its all delivered just before I get back home.
I think I have everything I need. Thanks everyone.
