The reports of combat robots headed for Iraq are somewhat misleading. These UGVs (Unmanned Ground Vehicles) are controlled remotely by a human operator. The weapon these UGVs will carry will be fired by the human operator. A truly robotic weapon makes its own decision when to fire. UGVs that operate like this are not far away, even though thats a scary prospect for many people. But such true combat robots are nothing new, they have been around for over a century. The first such combat robots were naval mines. Which lurked just below the surface, and exploded if a ship ran into them. Simple, effective and definitly robotic. Another 19th century development was the torpedo, which was initially thought of as a self-propelled naval mine. That was an accurate description, until World War II. That was when even smarter torpedoes were developed, like the German acoustic homing torpedo. This one went after the sound of a surface ship. After World War II, the wake homing torpedo (which sensed the wake of a surface ship and used that to find and destroy the ship) was developed. Land mines are also robotic weapons, and new versions were developed in the 1980s (WAAM) that were placed on the surface, and used sensors to detect targets. Anti-vehicle weapons were then fired into the air, and used more sensors to find, identify and destroy vehicles below. There are now naval mines with powerful computers on board, which make complex decisions on when, and what, to attack. These robotic weapons don't attract much attention.

But self-propelled armed land robots will not be so easily accepted. And we will soon find out to what extent. For example, there are already heat imaging systems equipped with software that can identify vehicles and people. These are now used for security purposes, but the technology exists to fit a weapon to these systems and produce a self-propelled vehicle that can patrol an area, and shoot at any vehicle or people it encounters. Manufacturers of these systems will point out that friendly troops operating near the killer robots can carry IFF (Identify, Friend of Foe) devices. But several generations of science fiction stories featuring evil killer robots will make these systems controversial. In response, developers are proposing fitting such systems with non-lethal weapons. A more likely solution is making the systems semi-autonomous. That is, the robots have to call for a human operator to confirm the use of lethal force. However, as the software and sensors become more capable, there will be the temptation to let the armed robots act on their own more and more. This will come from the fact that the battlefield is becoming more and more automated. This reduces friendly casualties. But the troops who are out there are always overworked and short of sleep. Letting the robots stay up all night pulling guard duty is an ever-present temptation. One solution for that is remote support. The technology exists to have some of the armed robots monitored via satellite link by troops, or even civilians, somewhere else in the world, like back in the United States. This is already being done with some UAVs. 

In theory, the very intelligent and capable armed robot is possible. But human nature being what it is, people will remain somewhere in the loop for some time to come. Or maybe not. Future battlefields will be full of armed robots, exactly how these robots will operate remains an open question.