Move over, Buck Rogers. Soon there may be a war-time battlefield where nary a human combatant is in view, but one in which swarms of unmanned, unattended, and untethered drones on the ground, in the air, and underwater are doing everything that is normally seen in a hostile combat zone: surveillance, strike, perhaps even capture and detention. Imagine hundreds, or thousands of vehicles – perhaps robotic flies or birds – all armed and dangerous, and, best of all, all doing the dirty work.
Now imagine this network as mobile – made ready in a matter of minutes – and modified for use in a massive search and rescue scenario, the likes of which we saw for instance, sadly, on September 11th.
This is the aim of Allen Moshfegh at the Office of Naval Research. His envisioned scenario goes beyond autopilot. It goes beyond preprogramming. His vision involves a ‘commander’ giving a broad operational order to thousands of drones and other vehicles – or agents as he likes to call them – but does not require that same commander having to manage all the minute details of the particular operation he has ordered. In fact, the agents will know exactly what to do, and just how to do it even if things go terribly wrong, as they so often do in war.
Moshfegh has pulled together a remarkable collaborative team from academia (notably Mario Gerla and John Villasenor of UCLA’s School of Engineering and Applied Science), industry, and federal government to develop his Autonomous Intelligent Network and Systems (AINS) initiative. His team includes electrical and chemical engineers, biochemists, and computer specialists, as well as neurobiologists and cognitive neuroscientists. Why these latter? Because Moshfegh sees the similarities and parallels between the way our brains convert decision to intent to action – not neuron by neuron, but only as a coordinated effort between hundreds of neurons. He wants to get his agents to perform in much the same way.
Networking a very few autonomous vehicles and performing an operation successfully is already possible. Moshfegh’s intent differs in that it involves whole battalions of these vehicles working and communicating and operating together. But, this can only happen if these vehicles can, after a fashion, think for themselves, and if the inherently prickly problems of such a system can be overcome.
What are those prickly problems? Consider mission planning and decision making with what is essentially an army of brainless hardware. Then include distributed signal processing, stored sensory information, distributed computing, mobile internetworking and wireless communication, monitor and control, variance in available bandwidth, hardware engineering difficulties, fault tolerance (the adaptive ability to protect and heal itself from the environment as well as from attack), not to mention demanding wireless communication networks (what Allen calls “Internet in the Sky”) needed by all these vehicles (some moving at 300 mph).
“As always, the ‘devil’s in the details,’” notes Moshfegh. “But this will work. We’ve seen it on a small scale already in the war in Afghanistan.”
An overview of the current status of the AINS initiative was presented earlier this month at UCLA. Copies of the scientific papers presented are available to the press.