I would like to build a 6+ legged robot that I can make to operate autonomously, probably a high target to aim for for a first project but what the hell. I have both electronics and computer training from previous jobs. But wher to start? What book to look in? Where to get parts? What parts to use? I have access to old and new computers if these are needed for the 'brains' but how do i go abount building/buying interfaces between the brain and the legs/sensors.
That's actually not too hard, depending on what you want it to autonomously do. If you want it to just walk around and avoid bumping into things, that's pretty easy. If you want it to serve drinks at your next party, that's harder. :)
I wouldn't suggest trying to use those computers, except for programming the microcontroller that you'll actually use as the brains on the bot. I'd suggest starting with a hexapod kit -- Lynxmotion has some nice ones, and the ones based on their "Servo Erector Set" are especially cool in my eyes because if you get tired of your hexapod, you can take it apart and build something completely different.
Start by building it and speaking directly to a servo controller (the SSC-32 is a good one) from your PC. Then choose a microcontroller (there are gazillions to choose from), program it in your favorite language, and stick it on the robot.
For sensors, you'll probably want at least a few of the Sharp IR distance sensors. These are available in various places, such as the Mark III Robot Store:
As one of the gazillion microcontrollers suppliers out there, I'll suggest you buy one of our ServoPod(TM)s or IsoPod(TM)s and save your money on the servo controller, because our micro is so advanced it can drive up to 26 servos from internal hardware, without taxing the internal processor. Which means with our micro, you can set an RC Servo, and forget it. The hardware will continue to generate the position pulses with no intervention from the software.
You might look at the various videos on Lynxmotion site, to see our software running hexapods, dancing sideways and diagonals, etc. Those videos are kind of impressive. Anybody else ever seen a hexapod walk on a diagonal?
-- Randy M. Dumse
formatting link
Caution: Objects in mirror are more confused than they appear.
STage 2 is to get it to learn about its environment and remember paths between certain points, the hexapod looks good, would I be able to evolve it to do the above?
I suppose I need to know if i can add extra sensors and some sort of memory functions to it either programatically or via hardware.
Difficult part is it takes full kinematics to do properly. Don't think most people bother. So you can draw your feet on a straight line in any direction? Do you have videos we could see?
-- Randy M. Dumse
formatting link
Caution: Objects in mirror are more confused than they appear.
Right now the software I'm using isn't perfectly precise. Good to maybe
+-5 degrees accuracy.
I have written some software that will give me +-.5mm accuracy in leg movement (and thus a fraction of a degree accuracy), but I am still testing it.
I still fail to see what is so terribly difficult. All you have to do is implement the full reverse kinematics for the legs. Basic trigonometry assuming a 3-DOF leg. Anybody that got through highschool math should not have trouble with it.
I don't have any videos of it walking in a non straight direction, sorry.
What you say is true, and particularly for someone for instance from the Physics department (refering to myself). However, if you check with University level students, and ask them about the ease of kinematics, you'll get quite a different story. Full kinematics is "a big deal" in the industrial tech department. Or at least that has been my impression. Besides Mike Keesling and myself, you're the first one I've ever heard that has done it for a hex 3dof walker. Ask around, and see if what you find is different.
-- Randy M. Dumse
formatting link
Caution: Objects in mirror are more confused than they appear.
Hmm - I figured out the kinematics for my robot before I knew what kinematics meant. I think most anybody capable of building a 6 legged robot should be able to do it. I think one thing that may prevent many from implementing full kinematics is the processing power necessary to compute the positions of all the motors in real time if you want smooth motion. And just laziness, of course. There are all sorts of approximations that can be used - but for perfectly smooth movement - you need a decent amount of processing power, at least I did.
Anyways - I'm glad I'm among the elite few. If you're interested, I'll post a video of him walking at a random angle when I get my new control algorithm finished (should be about a month or two). Hopefully the new algorithm should look pretty darn smooth. We shall see!
There's no doubt it takes a fair amount of processing power to work all the trig. functions working through the kinematics. Particularly when you're considering 18 servos, each with a different angle, and doing it over 50 times a second.
Game developers, however, do it all the time. Look around for inverse kinematics code. For legs with unique inverse kinematics that don't operate near their singularities, it's not that bad.
PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.