Let's Make Robots!

To do list!

I don't really know what this is for - so apologies if it's not what I think it is and my rambling are cluttering up front space in anything/something.This won't make very interesting reading, but it's to document where I'm up to a bit.


So here I'm putting my unformed ideas, with the intent to figure out how to make them exist.

Great Idea Number 0; pcb scanner/engraver.

 The genesis of this idea is my present inability to make a CNC machine - or at least, figure out how to interface it with a computer. So I want an engraver that I dont HAVE to interface with a computer and can just figure out what it needs to do from a drawing on a piece of paper. The only way I can think of making this work is as a kind of simultaneous single-pixel scanner and single-pixel dot-matrix printer...

To the present idea is to have a small (say 200mmx200mm) two-level tower. Mounted on the tower is a rig with 2-D freedom in x and y, along with driver motors etc. etc. . I figure that I don't need stepper motors, since the things pcbs I'll be engraving will be around 100-150mm at most, and over that size a servo has about 1mm resolution, which is all I need for a pcb.
To do; check the stability of servo calls - i.e. if I command servo 1, 75 then servo 1, 200, then servo 1,75   - has the servo returned to the same place as it was in the first call?

On the top level of the tower, the rig holds a dremmel. The dremmel will need some short play in the z direction, so it can be driven up and down a short distance.Engraving resolution will be 1mm so the dremmel should be able to engrave a line that wide. Assuming that the machine will engrave at a rate of about 10mm per second, it will take just over 15 minutes to engrave a 100x100 mm board into oblivion (i.e. totaly) a 5x5 mm board will take about 4 minutes. Great! I spend at least 15 minutes etching..
To do; Check dremmels and their performance. Is 1mm resolution realisitc AND practical? is 10mm per second realistic?
Result; um.. from a ¥1000 dremmel, the answer is no. then again, it's a pretty poor excuse for a dremmel. Think of something else!

On the bottom level will be where the paper drawing is scanned. Scanning resolution will be 1mm. A high powered-led will shine thru the paper on the bottom, and be scanned by the 2-D rig. In this case, a small (3x3mm) optical fibre array - the tips of the fibres will sweep over the paper, and light will be relatively darger as the fibres pass over a line. The signals are processed and transmitted live to the dremmel on the level above - light means engrave, dark means do not. Having an array of optical fibre tips means that the engrave status is known before it's actually engraved.

edit - Bummer; I bought some "optical fibre" from tokyo hands, that turns out to be just brittle acrylic strands. Total shite.
edit - SCOREl I found a whole lot of optical fibre at uni that was about to be thrown away! super score! This is really thin though- group in bunches of 3-4 strands.

To do; test the transmission of optical light thru paper, down fibre, into a (pricy! .... for me) TPS603A phototransistor.
Result; Thru paper transmission is lousy- better idea is to aim laser onto page surface position fibre top on laser spot. Track the laser and fibre together over the page.  This seems to work - I had a 5% defletcion of light transmitted down fibre when the fibre+laser spot passed over a black line on the paper. Ganged trannies on phototrannie output to get a useful signal. Probably should get a darlington ic. hell, get a few.

Note; optical fibre melts VERY easily - to interface fibre with PT, contain the fibre bunch with alfoil, then put a bead of hotglue on the tip of the PT and let it cool a bit. Insert the Fibre tips (wriggle it around so that there is a good contact with the PT) and hold till the glue hardens. Then slide some heat-shrink over assembly and, using more alfoil to protect the fibre, heat the heatshrink with the flamethrower.

Get low-cost laser set from Kamimaizu (get one that is well-focused). I just smoked my old cheap(er) one.

Great Idea Number 1; Counter balls.

So this plan is borne because it's a great idea (IMO), and I think there are enough bits and pieces around the place to put it together. My intention is to have transparent ball shells, with a radio transmitter, receiver .. - and a POV device capable of displaying numbers around the inside surface of the ball.

So one counterball by itself will send out a very weak "I'm here" signal. Any other counterballs within range will respond, and transmit their ID  - some code number that I will make up. Ball one will compile a list of who is nearby, and RETRANSMIT that to it's mates. It's mates will compare THAT list, with their OWN list. If they match, or mostly match, then they can decide to be a group, add up the numbers, and all members of that group will display that number via POV.  Each member of the group does this at whatever rate is possible (but quite quickly, I hope - no more than 5 seconds!).

Taking one ball out of the group and putting it elswhere will  mean that group has lost a member - which they will discover, and they will all decrement their number and redisplay the new one.

Similarly, adding a new one will cause the rest of the group to increment their number.

Having small clusters of balls together will form separate groups, each ball in each group will show the number of balls in their group, and continually update it via radio.

The function is really just to provide a very visual idea of, for example, what "3" is, and how it's made up of things like "2" and "1". Sounds stupid? no way, it's awsome and you'd be so wrong if you ever thought that! - it would be a killer idea for young kids - complete with pretty flashy lights and zippy crashy sounds.

Problems/difficulties etc.

1. Well, I really have no idea how to do most of this, but the best thing to do first would be to figure out how to transmit that info between ball groups, and write code for the balls to figure out which number they need to be.

2. POV -I have yet to figue that out too, but lots of people do it, so it can't be imposible.

3. Conservation of Angular momentum - this is a beastie that pops up often with spinning things - especially affecting things that only have one point of contact with the ground, like a ball. The solution might be to have a small counterweight that spins in the opposite direction or something, but it's going to be strange to find out how best to mitigate it....



Great Idea Number 2; A weighted companion cube roomba.
If you don't know what a weighted companion cube is, then I suggest you google "portal", clear your schedule for half the weekend, and get busy...

 I am in the middle of making a vacuum bug, to slurp up gritty bits, but I think a cube shape roomba is better. So I'll be trying to make one of these in the not-amazingly-distant-future. It wil be a square roomba, about 7cm high, 23cm on a side, and look like a piece of a WCC. On each side, there are two trigger switches, mounted on the side, right near the corner, so they can be activated if it hits a wall.

The cleaning algorithm is simple: 

0. turn on vacuum.

1. Continue  straight ahead, or randomly deviate from straight ahead, until a wall is hit - this will be hit at an angle (one trigger switch is ALWAYS triggered before the other..)

2. stop and make a rotation so that the triggered corner rotates AWAY from the wall (and conversely, the untriggered switch on the same side rotates towards the wall) and count how long it takes to get there. Rotate BACK the other way, half that time, and then the edge of the WCCR is paralell with the wall edge.

3. continue in one of three directions - along the wall left, right, or even back away from the wall.

Randomness is key to a vaccum cleaner, but walls should be given special attention.



Great Idea Number 3; A ball drive?.

Well, I don't know what the best word for this is, but I've seen them around on the internets, and I thought it would be pretty straightforwad and cheap to make one ... famous last words.


The idea is to have two orthogonal motors that contact a rubber ball with metal bars. The ball is held more or less in place. The two motors can rotate the ball around any axis EXCEPT the vertical axis. 

The two motors can be servos, as in the pictures




Or for smaller size, add an ungeared motor that worm drive

What's the point?An excellent question, but I was inspired by patrick's chess computer, and I thought it would be neat to make pieces that could move themselves, under the control of a central brain. Come to think of it, it might have also been a good xmos idea, but alas, I have no time....





Other ideas that I'm too lazy to detail at this point..


two-footed weight transfer stepping robot. Mounted on a rocker and with a trolly that runs on rails between the feet, to transfer weight.

Neko busu (cat bus!)

quad stepping robot.

bipedal robot with two oppositely swinging pendula to acccompiish weight transfer.




Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

I like the first idea. You could get plastic balls from those gatcha-gatcha machines down at the game arcades. Once you get the toy out of the plastic ball most people put the balls back into a collection box. Just grab a few empties when nobody is looking :D

Forgot to mention that Xbee supports multipoint networks for $22 each.