Students at the University of
Southampton have developed a low cost robot
for swarm intelligence research.
"The clever issue is using surface mount motors from mobile
phone vibrators so the whole thing can be made on a standard
production line," academic Klaus-Peter Zauner told Electronics
Weekly. "There are no special components and no hand assembly."
Estimated price of the 30x28x12mm machine is £24.
Zauner was at pains to point out he had nothing to do with the
initial idea. "Students built the prototype by taking old mobile
phones apart," he said. "I was so impressed that I arranged some
funding."
Following natural examples including bees and ants, swam
intelligence research attempts to discover 'emergent behaviour -
where individuals following simple rules work together to produce
complex results.
The whole process can be modelled in computers - there is a
commercial spam email detection system that uses communicating
software 'agents' - but sometimes the is no substitute for tens,
hundreds, or even thousands of physical robots.
With an eye on cost, component count has been kept to a
minimum.
The robots two surface-mount motors have wheels on their
spindles and are soldered directly to the underside of the
double-sided PCB which is also the robot's chassis.
Power comes from under-slung 320mAh Li-ion cell sufficient for
1.5h of operation between charges.
For their size, the robots move quickly: EW estimates they
travel at somewhere between 300 and 500mm/s.
On-board intelligence comes from a 16MHz MSP430 microcontroller
from Texas Instruments with 16kbyte flash and 512byte RAM. Of this,
6kbyte flash and 256byte RAM are needed to work the machine, the
rest is available for behavioural algorithms
Three fast infrared sensors (black in the photos), mounted on
top of the PCB facing horizontally at 120[deg] intervals, allow
data to be received from other robots in any direction.
Three infra-red transmitters (white), mounted evenly between the
receivers complete the 588bit/s broadcast-only communication
link.
According to Zauner, the transmitters, and the receivers, are
operated in parallel, during communication. Range is deliberately
limited to around 10cm so robots communicate only with immediate
neighbours.
IR communication allows not only co-operative behaviour but, as
the robots can flash their own programme memory, allows new
operating code to be passed around.
The protocol for re-programming includes a re-transmission
request for robots that miss a code segment.
The IR transmitters and receivers can also be used as a simple
wall detector. "When they start receiving their own packets, they
know they are bouncing off something nearby," said Zauner.
Lastly, the three receivers can also be used to estimate the
direction of bright light sources in their environment.
The only other sensor is a infra-red transmitter-receiver pair
mounted underneath at the front which allows the robot to tell if
it is over a dark or light surface.
The university's robot experiment arena has black wallpaper on
the floor. "Someone took a robot to a DIY store and found the paper
that reflected least," said Zauner.
One of the experimental objects made by the team is a balsa wood
puck the size of a 2p coin.
It is stuck to a slightly larger circular piece of white paper
and the geometry is such that the dark-light sensor can only see
the white paper when the robot is nose-on to the puck.
Using the three top IR sensors as directional light detectors,
and the underneath sensor, rules have been developed that cause a
group of the robots to nudge this puck away from a bright light at
the edge of the arena.
The robots know when they need to be charged and know where
their charger is.
Four bare copper wires at the rear of the chassis, two for power
and two for ground, pick up current for recharging. The two ground
wires are bent so they double as skids in lieu of wheels.
Aside from the charging circuit, the only other components are a
handful of drive transistors and a bi-colour LED on the top to
indicate activity to human bystanders - a cluster of green LEDs in
the arena, for example, indicating the swarm has found something of
interest.
Zauner and the students: Alexis Johnson, Stephen English,
Jeffrey Gough, Robert Spanton and Joanna Sun are part of the
University's newly-formed
Science and
Engineering of Natural Systems (SENSe)
group within its School of Electronics and Computer Science
(ECS).