TreadBot

[skimask]

the mapping idea is just for the challenge of trying it, but im still debating my method of using the revolutions to tell how far its gone.. the room (the main floor of the house) has linoleum, hardwood and a couple rugs.. i can almost guarantee its going to slide trying to go from one to the other.. it looks like i may need to add a compass into it somehow, or add IR transmitters on different frequencies around the room, and have the robot detect them and figure out where it is...

im curious as to what other people think about what it can do, or how to figure out where it is...

Mapping - are you using stepper motors or regular 'just add volts and it spins' motors? Regular motors - good luck getting them to spin EXACTLY the same speed to go straight! Not saying it's impossible, but it'll be tough without good electronics. And what happens when the treads slip on the floor? The 'bot could think it's moving forward, when it's actually getting nowhere. Keep an eye on current draw could help with that. Motor draws X current, it's moving normally, motor draws 2X current, tracks are probably hanging up, motor draws .5X current, tracks are probably slipping like they're on ice.

Compass - magnetic compass 'could' work. How about an RF/IR link? A master RF TX sends out a pulse at X intervals, say 10 seconds. 1 second later, an IR LED on wall X1 pulses, 1 second later, an IR LED on wall X2 pulses, and so on and so on. The 'bot 'hears' the master pulse and starts looking for IR pulses. Depending on which IR sensor sees it and at what time, it can figure out which way it's pointing, limited accuracy of course.

[dragons_fire]

the motors are just regular brushed motors, they have a gear ratio of 298:1 which equals out to 71RPM and 30oz/in of torque. im using a serial dual motor controller, and with the hall sensors, i should be able to adjust the motors to get it to move straight forward...

my big issue, is that it will turn better on shiny hardwood, than it will on carpet, so if i tell it to turn 90*, it will turn 95* on wood, and 80* on carpet.. what i was thinking with the IR link, was to have 3 transmitters around the room, and then have 16 (or more) sensors around the top of the bot, and it could somewhat triangulate where it is by the angles to all 3 transmitters. it could do that maybe every 5 minutes, just to make sure its where its supposed to be...

[skimask]

the motors are just regular brushed motors, they have a gear ratio of 298:1 which equals out to 71RPM and 30oz/in of torque. im using a serial dual motor controller, and with the hall sensors, i should be able to adjust the motors to get it to move straight forward...

my big issue, is that it will turn better on shiny hardwood, than it will on carpet, so if i tell it to turn 90*, it will turn 95* on wood, and 80* on carpet.. what i was thinking with the IR link, was to have 3 transmitters around the room, and then have 16 (or more) sensors around the top of the bot, and it could somewhat triangulate where it is by the angles to all 3 transmitters. it could do that maybe every 5 minutes, just to make sure its where its supposed to be...

Motors - That kind of gear ratio sounds a bunch better than direct drive. Should be able to keep track of that fairly well.

Turning/mapping/whatever - too bad a guy couldn't 'easily' add a circuit like an optical mouse. Ya get 2 of those, one at the front in the middle, one at the back in the middle, you can do some math to figure out how far the 'bot actually turned.

This could turn out to be a lllooonnnggg thread...

[dragons_fire]

definately could be a long thread... there are a bunch of different ways to do everything... like figuring out how far it turns... there is: gyros, compass, "mouse balls", hall sensors on the wheels, IR base stations (celestial navigation), and im sure there has to be another 100 ways that i didnt mention...

right now im thinking i might just make it so it hides from light (photophobic) during the day, and then comes out to play at night... at least until i can come up with something better for it to do...

[skimask]

right now im thinking i might just make it so it hides from light (photophobic) during the day, and then comes out to play at night... at least until i can come up with something better for it to do...

Just like in Gremlins...(you said 'mouse balls').

[dragons_fire]

HAHA i havent seen that movie in forever....

[mackrackit]

Turning/mapping/whatever - too bad a guy couldn't 'easily' add a circuit like an optical mouse. Ya get 2 of those, one at the front in the middle, one at the back in the middle, you can do some math to figure out how far the 'bot actually turned.

Why not use a "ball" mouse. xy axis could keep track of turns to. Might take a bigger ball to go across the floor. The optos in the mouse are easy to use too.

[dragons_fire]

i didnt know computer mouses had balls??? i thought they all used lights and lasers!!!!! :P

i like the idea, but i think it would have problems getting up from the hardwood to the rug, (about 1/4" bump) and still be somewhat accurate with out using a massive ball, and i dont really have much room for a big ball...


Not trying to put down all the ideas, im just trying to get some ideas and start a conversation about different ways to do it... its really interesting how many ways there are to do different tasks and some of the weird ways that no one would ever think of.

[mackrackit]

"i didnt know computer mouses had balls???"

That is how you tell the difference, male or female... You really are a NEWBIE

The ball and optos could mount on a spring loaded platform in the front or back.

This is an interesting problem, in my work we use doppler radar along with gps for applicators that move across the field at 20 mph. Even then things could end up depending on plowed or disked - satellite coverage - etc- 50 cm off in a couple of km.

Run a "female" mouse over paper with lines and you will see the problems with that. I bet the RF idea with beacons would be the most accurate.

If I think of something very small and cheap will let you know.

[skimask]

Awhile back I had an idea for tracking a push type lawn mower.
I'd put IR LED transmitters in the 4 corners of a square lot, each with it's own pulsing IR code. On the top of the mower would be an IR detector that could move 360 degrees. I slowly go 'round until I saw 1 of the IR LEDs, mark that angle, and keep going until I saw at least 2 of the 4 IR transmitters. As long as I had two angles to work with, I could triangulate my position inside the square lot (ignoring the fact that there are really 2 places an object could be in 2D space using that method, one inside the square, one outside). The more the IR LED transmitters, the better the accuracy, and the better the chances I'll see them if the platform was to somehow get off level.