Slashdot Mirror
Armed Robots Not Actually Gone From Iraq
NightFalcon90909 writes "You may have heard that armed robots were yanked from Iraq after a gun started to swivel without it being told to do so. 'A recent news report that armed robots had been pulled out of Iraq is mistaken, according to the company that makes the robot [Foster-Miller] and the Army program manager. 'The whole thing is an urban legend,' says Foster Miller spokesperson Cynthia Black, of the reports about SWORDS moving its gun without a command.'"
I'm an engineer for a company that writes some of the signal analysis for robots, mostly military. They are designed to look for people, noise, or something easily sensible and train their guns on that location and await further instruction. Its a de facto law for military robot design that a human makes every firing decision, but the robot is allowed to aim and ask if it can fire. If a US soldier did something loud (shoot a gun, slam a door, yell) theres a good chance thats what set off the targeting routine. There was never any chance of a weapon being fired, except of course if there was a malicious operator. I have not worked on this type of robot, so I can't be sure of the process. There might be a user command that says "go look for target". If the robot looked for a target without ever being commanded that'd be a pretty horrendous software error.
Disclaimer: I haven't worked on SWORD robots, but I have worked with the TALON on which the SWORDS are based.
The sort of scenario you describe is prevented with a heartbeat based killswitch. E.g. a signal is sent to the robot at a regular interval. If, for some reason, the heartbeat is not received, the robot immediately shuts down and stops moving. So, as you said, the robot "stops cold any time the transmission is having a hiccup." It can be a pain sometimes, but it's hell of a lot better than the alternative.
In the same way, dangerous commands (such as "shoot gun") require the robot to receive said command constantly in order to continue that action. So a robot being commanded to turn and fire just before losing comms would at worst, just turn, and typically do nothing.
Also: +1 Ironic Sig.
Any plan which depends on a fundamental change in human behavior is doomed from the start.
if the bit that makes it swivel engages without being told, what on Earth makes you so confident that the bit that makes it shoot will not engage without it being told? To answer your question, not a damn thing. The TALON I worked with was really flaky. It shook and twitched so frequently the guys who owned the TALON referred to the bot has having the "Foster-Miller shakes."
I hope the SWORD bots are much better quality than the TALON bot, because, quite frankly, there is no fraking way I'd trust one of those things with a gun.
Any plan which depends on a fundamental change in human behavior is doomed from the start.
Sounds fishy to me.
None of the ships involved in the initial Aegis tests can be described as "automated vessels". The initial radar tests were aboard USS Norton Sound, later tests would have been on USS Ticonderoga. Neither use Windows NT, and in neither ship was/is the Aegis system connected to the propulsion or navigation. Pulling the plug to the point where the ship was dead in the water wouldn't have been necessary on either.
Also, there is no "Aegis Class Cruiser". The Ticonderoga class cruisers use the Aegis combat system, but so do several other ship classes (Arleigh Burke, some Japanese and Spanish ships as well).
There was an incident where an experimental Windows-based ship management system (again, separate from the combat system) caused a Ticonderoga-class ship to lose propulsion.
In other words, the part that says illegal combatants STILL HAVE RIGHTS, and the right to a trial is explicitly mentioned.