AUVSI

[Noel Sharkey]

The UK's Daily Telegraph reported today (April 23) that one of the UK's 3 MQ-9 Reaper drones AKA the Predator B crash landed in Afghanistan on April 9 this year. Although the Reaper can carry 14 Hellfire missiles, the UK Reapers are not armed yet.
http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2008/04/22/nafghan222.xml

What is important to note is in how quickly British Special Forces moved in
to strip the wrecked plane of its camera and memory chips before a manned Harrier dropped a 1,000lb bomb on it.

The Telegraph article reports

[Luke Haberkern]

The self-destruct concept is already used in aerial bombs, mine detonators, and various other systems. Why then, can we not use such systems with autonomous and semi-autonomous systems where the weapons support crew would have the sole decision when such actions are necessary. Obviously we cannot have the weapon make its own choice in this for fear of accidental detonation (especially) around fellow soldiers. The aspect of losing such information to the enemy are so devastating, however, that it is necessary for such devices be used in "last choice" scenarios. What do you think?
-Luke

[Noel Sharkey]

John Canning would be better at answering this one but I will have a go.
Luke raises a very good point here.

My information from a number of sources is that self-destruct has been considered for Unmanned Aerial Vehicles like the Predator MQ-1 and the Reaper MQ-9. I was decided that this was not a good idea because of possible civilian or friendly deaths that could result. Don't forget that a vehicle might have received damage to its sensing equipment and not be able to furnish information about its whereabout or who else is in the vicinity.

A recent example is when the UK owned Reaper crashed - special forces went in to remove the hard drive with the reconnaissance info on it and then the Air Force dropped a 1,000 pound bomb on it.

Of course another issue is that because of the huge cost of autonomous ground systems, self-destruct rather than self-defense would be a very expensive option. If the harmless self-destruct was possible with say small charges to destroy the essentials and not harm others outside the vehicle, then it would be an ethically better option.

I would love to hear what others think about the options.

[John Canning]

The issue with incorporating a self-destruct capability is wrapped-up in the issue of limiting collateral damage.

We talked to the JAG lawyers about this several years ago and were told not to even _think_ about going there! They told us that this leads into the same area as the issues with anti-personnel landmines.

A self-destruct capability, however, is not necessarily the same as an anti-tamper capability, but you can achieve many of the same goals with good anti-tamper mechanisms as you could with self-destruct capabilities, but without all the messiness involved with destroyed parts flying around. You just need to think about these a bit more carefully when you design your robot.

[Luke Haberkern]

I apologize about being unclear, but when I mean self destruct, I had no intention of implying an actual explosion going of in or around the device. The self destruct mechanisms I have worked with are simply a an electric switch that forces a lot of current backwards through the circuitry of the device effectively "frying" (though I hate that term) every component within it. The debate, then, is not whether its annihilation could hurt others, but if the information that can be pulled from the robotic corpse is enough to assist the enemy in their own designs. Again, I am sorry for not being more clear.
-Luke

[John Canning]

Ah! Now I better understand where you are coming from on this, but your concern is something that a good anti-tamper design should address. There are ways of dealing with such issues, but I won't address them here.