I actually work on the team that handles the swarming aspect of the system. Entities in the swarm do not report to a central server as you purport. It is not PSO which actually has nothing to do with our approach. Each entity is responsible for knowing the location and task of other entities in the swarm. This enables us to scale the system very efficiently by eliminating a centralized server that maintains knowledge of the environment and entity state.
While the swarm is autonomous, there can be a man in the loop. In the scenarios we were testing in, this was a requirement but that is not to say that it couldn't run without one. Even when operational, I do not see there ever being a case where a human operator will not be in the loop given current doctrine and rules of engagement.
You are correct, this is not really "terminator type stuff," it is an effort to decrease human risk, increase intelligence gathering and other operational efficiency.
With so may references to Skynet in this entire thread, if this truly is the start of Skynet, then I for one will welcome my Terminator assassin sent back from the future to prevent this from happening and you can all say that you told me so.