Quadrotors learning to throw and catch
mashable.com
mashable.com
Besides the obvious manhack reference from Half-Life 2, this sort of cooperative robotics seems like it will help solve problems that single large robots cannot.
For example, the researchers could use their tool box of behaviors to create a swarm of robots which create a 'bucket brigade' from a source of water to a wildfire. If the only limitation on the rate of water drop is number of robots in the brigade, it becomes a simple economic question rather than a policy question as to whether or not to implement something like this. Given the cost/damage of wild fires it seems like a case could be made for a few thousand gas powered water bucket carrying quadrotors.
My reasoning is that they would respond more rapidly to changes in atmospheric conditions. They have the advantage as well that if you lose one you don't lose multiple thousands of gallons of water/retardant capacity.
Better use electrically powered ones (though the range may be prohibitive).