A draft scenario for NEST, from Greg Sullivan
Demonstration project for NEST:
Mission: distribute vision and motion sensing devices to a scene and
monitor it.
Scenario:
1. A herd of wheeled teenies, some with video and some with motion
sensors, are assembled around a larger RC car. The RC car is
itself equipped with a camera and is in radio contact with the
home base and with its teenies.
2. The teenies are programmed to herd.
3. The horde follows the car, demonstrating herding. At various
points, teenies drop off to maintain networking waystations for
short-hop networking. This dropping off and going into network
waystation mode is probably directed from the RC car, via the
base.
4. When the herd reaches its destination (probably the 8ai
playroom), the RC car directs various teenies to their monitoring
stations and gives (downloads to) the teenies their monitoring
programs.
5. The RC car returns to base.
The monitoring phase consists of:
* when motion sensors trip, they wake up the cameras.
* camera teenies spend some time acquiring images, getting average
luminosity, and then communicating the average to base.
* along the route to the base, an optimization is employed: when
two sources are each sending an average luminosity, their averages
are averaged (and weight increased) and only one message is passed
further along.
* when the base receives all (or most of) the averages, a global
average is computed. Note that if there are enough hops, it might
be that the base receives only one already-calculated average.
* Each video teenie is then instructed to monitor with respect to
the average luminosity. That is, images are compressed with
respect to the average into light and dark, with segments
identified using lines. Note that when all of the teenies are
sent the same information from the base, only one packet of
information is sent, and it is distributed to each of the video
teenies (as opposed to addressing and sending to each teenie
individually).
Carrying on:
* monitoring, analysis, and adaptation. Reconfiguring the herd to
do a better job. This might involve both physical reconfiguration
and software reconfiguration.
Highlights:
* multiple levels of control.
* dynamic reprogramming of teenies.
* optimization w.r.t. communication cost.
Options:
* maybe microphones also?
* monitoring teenies? Some of the teenies are "monitor" teenies --
that is, their purpose is to monitor the behavior of the other
teenies and provide monitoring, diagnosis, and repair services to
the system as a whole. At a minimum, the monitor teenies would
notice when teenies "go away" and notify the base.