A wired blog article broke news about the Defense Advanced Research Projects Agency (DARPA) newest public project. These little pals are robotic wireless access points, or more properly - autonomous, mesh networked, wireless extension units. The agency sees the future of the glorified lego mindstorms robots as being expendable units for soldiers to deploy as they move, thereby extending the multitudinous information technologies available to field commandos. Each robot must be less than $100 per unit on small production runs, must have a long service life of 7 - 14 days while on batteries (they won't be recharged), and they must be intelligent. Swarm technologies have been used in robot demonstrations where each unit is aware of each other and actually work together by passing information from unit to unit to achieve the objective. These units would be able to heal the network, should an individual unit shutdown or be destroyed.
I found myself wondering about these technologies. I feel that the most difficult requirement to satisfy will be the cost measure. I'm sure that some University has been working on just such a project somewhere but otherwise, how could a corporation hope to make and $$ on this? There won't be any support because they're throw away and self programming. There won't be any parts market for the same reason so all of their costs for production and development must be paid off within the sale price. Of course, like every government contract, this project will be going to the lowest bidder.
The hardest part of the hardware will be designing a router that will interface with both the data and voice networks of the military (NIPRnet and SIPRnet plus many others that were not declassified at my time in the Air Force) while at the same time being able to communicate with the other units in the area. The truss and chassis will probably be a lightweight metal with treads of some sort. The battery unit will invariably be the largest portion of the entire unit since it will need to power both the propulsion unit and the router technology. But these hardware issues can surely utilize market technologies without needing in house development.
Software will be a different issue all together. Not only will the routers have to recognize where information is going and coming from, but they'll need to know where the other units are such that they can cross route information without having to go back to the main servers in order to facilitate faster data transfers. Each robot will need to have a broad understanding of the current network coverage as well as the holes in reception and be able to navigate to improve connection strength and safety. Fortunately, DARPA is looking for these units to fit into the urban warfare environment where coverage must be robust with respect to structures and shorter distances rather than the more open distances of desert or wilderness environments. Perhaps the military will integrate the GPS system into the unit providing waypoints for the robots to traverse but this would increase costs and incur further programming time for what was supposed to be a zero-configuration project.
In summary, I think that this particular project will be a new step for DARPA by showing their interest in disposable network technology (tolerance for it falling into enemy hands with no plan for recovery by US forces... can you say self destruct?) as well as their continued insurmountable cost positions. Let's get real... even if a company was going to get further gov't contracts as a result of this, there is no way they'll be able to fulfill the mission specs while simultaneously staying within the price point for each unit.