On 01/03/2016 12:10, Nick Forbes wrote:
> I think we possibly need to consider two projects, one being an all
> singing and dancing cluster robot with bleading edge technology that
> will probably cost 10s of £s and a separate absolutely stripped to the
> bones one that could be assembled at open days etc. that would cost
> between £5 and £10 each.
Agreed. Our focus last night was on the cheap ones with the explicit
intention of getting people to make them and add them to the NADHack
cluster, i.e. not take them away.
> My interests are in robotics so I want one (at least) of each!
Same here :D
> Personally I think initially we should concentrate on a very simple,
> stripped to the bones (i.e. cheap) robot with very basic
> functionality, my first thoughts are:
> - Use pager vibrators and three legs to save on drive motors, gear
> boxes etc.
My biggest concern about this approach is that the play area needs to be
absolutely level, otherwise the robots will follow gravity. It is very
cheap, of course :)
http://www.aliexpress.com/item/FREE-SHIPPING-Vibration-Pager-Vibrating-Vibrator-Micro-mobile-Motor-6mm-x12mm-diameter/565241701.html
and
http://www.aliexpress.com/item/100pcs-lot-12MM-2-8-MM-Micro-Button-coin-pancake-Type-Vibration-Motor-2-V-5/32387232857.html
Both good examples of bulk buy vibration motors, about $30 for 100 or
$60 for 100 robots (2 motors each).
> - Rechargeable battery
18650 cells (much larger than we spoke about last night, about 65mm long
and 3Ah) are about 1.2USD each at 100 off.
16340 cells (16mm diameter, 34mm long and 1.2Ah) are about .9USD each at
100 off, or $90 for 100 robots.
> - Install a boot loader so we don't need fancy programming tools
Agreed. Using IR or RF for programming through the bootloader is a big
benefit, if we can keep the communication simple we can use the on-chip
UARTs and minimise the code footprint. In-protocol checksums are a good
way of ensuring that we don't corrupt anything.
> - Charge and program using the three legs to save on connectors, might
> also need a debug connector for initial development
Can definitely charge using the three legs (see below*) but if we can
program using IR (once the bootloader is on) then we can do contactless
reprogramming.
> - RGB led so we can make teams in the cluster or just indicate actions
> and they look cool
Definitely agree. I like the idea of having four (or eight, or
whatever) RGB LEDs to light the case for external robot recognition. If
we can recognise four colours then four LEDs gives us 4^4 or 256 unique
colours, but we'd lose 75% due to the rotational symmetry, i.e. we'd
only have 64 unique patterns. With eight LEDs in a ring and four
recognisable colours we can 8^4 / 8 = 8^3 or 16M unique robots. Putting
the LEDs in an asymmetric pattern (e.g. a U shape instead of a ring)
gives us better recognition (unique plus orientation) and you could do
1024 robots with six RGB LEDs, 256 robots with four RGB LEDs.
We talked about displaying patterns, e.g. a digital clock, so if you
used three robots per segment you'd need 21 robots to make a single
7-segment display. Clocks require approximately 4+two-halves segments,
so we'd need approx 100 robots to make a digital clock. If you forego
seconds (which the robots couldn't keep up with anyway) you can do it
with 2 + .5 + .5 segments, or approximately 64 robots.
> - IR communications - cheaper than radio
Agree that it's much cheaper but range and interference are more of a
problem. 100pcs of discrete IR receiver (e.g. TSOP1738) are $50 but
bizarrely cheaper from Farnell :) For the purposes of this email I'll
assume that transmitters are equivalently priced so 100 robots are $100.
> - Collision detection IR or possibly feeler switches
I'd love to do this but think this might be cost prohibitive. Using
infrared doubles the IR budget (but does allow us to use directional IR
for other purposes).
> - My personal preference would be to use a PIC as the processor as
> that is what I'm familiar with.
I personally like the AVRs but absolutely happy to choose any platform.
> Quantities - initial batch of 10 for dev work and then probably gear
> up for 100 for open days etc. however we wouldn't need to build them
> all at once. PCBs get cheaper by the 100, most of the other components
> only get significantly cheaper by the 1000 or more so we need to look
> at a sensible balance.
I think we could be in the region of $10 per robot at 100 robots, which
is an astonishingly low $1000 / £650. I'm happy to pledge a decent
chunk of that if it's something we as a group could realistically commit
to designing and manufacturing...
Stuart
*Footnote about charging... the charging area could be a lattice of
aluminium/copper electrodes with an AC power source. The legs could
then be fed through a diode rectifier to the charging circuit. For
swarmed feeding behaviours the charging area could be considered a food
source :)