Wow,
this is pretty damn cool!
Led by Dr. Owen Hollands, students in the University of Essex Computer Science and Electronic Systems Engineering Departments have a set of projects that are focused on building swarm networks. Essentially, swarm networks are self-configuring/self-aligning networks, designed to collect and process information in a distributed fashion. The real hope of swarm networks is to see emergent behavior arise from fairly simple rules governing control and response to stimulus. The current work these folks are focused on is creating air mobile swarms, but heterogeneous swarms would also be useful.
The two projects are named Gridswarm, and UltraSwarm.
The
Gridswarm project is based on UAV/RC airplane technology and does not appear to be substantially different from other projects I have seen. (Their
links page is a good resource for such projects.)
Except that they use a
WOT4. I believe this makes it a very cost-effective system compared to a lot of others I have seen. Obviously, they are not tied to this platform in the long run, but it seems like a nice system to start with, and is a great aerobatic model. It is powerful, and is stable at low speeds.
However, they do use a
MicroPilot autopilot system, which is not a cheap solution($1500).
Unfortunately, I’m not all that excited about airplane based swarms. Orienting a swarm of aircraft to collect information over a specific target, while avoiding collisions, seems very difficult to me, and not especially feasible. I don’t believe birds flock in order to process (or gather) large amounts of information. I
do see flocking as a possible method for rapid entry/exit of numbers of UAVs into a battlefield (especially urban environments). The emergent behaviors that result from flocking, such as collision avoidance and path finding, may enhace the target selection process, making it a good avenue of research. So perhaps airplane based swarms would be useful when modeling a bird like gulls, which flock in order to identify good feeding grounds, but are extremely competitive (and generally independent) hunters.
A interesting side note is that our current use of Predator drones is much more like a bird of prey than as a flocking bird. We position them relatively high over the battlefield and collect data until there is a target of opportunity.
Now the second project, UltraSwarm, is what interests me. They have used a
ProxFlyer Bladerunner with custom flight control (using a
Gumstix computer) to create a stable little indoor network node. I’d be really curious to see how it behaves.
Unfortunately, both projects have substantial problems. Power is a huge problem, especially for the Bladerunner. The networks are significantly limited by range and speed, due to the use of bluetooth. But I like both of them because they use a low cost approach.
The paper the group is presenting at the IEEE Swarm Intelligence Symposium is available at
http://cswww.essex.ac.uk/staff/owen/SIS2005copyright.pdf.
