Robotics and the future of warfare

I was speaking to one of my friends doing research in robotics in a major American university when it struck me that most of the researchers in robotics and machine learning are gentle pacifists who wouldn't want to hurt a fly, while one of the greatest uses of robots is going to be in real battlefields.

And I am not talking about robots engaging in non lethal activities like scouting or mine clearing, though that's how robots will initially appear on the battlefield. I am talking of robots (on land ,sea and air) armed with guns and other weaponry and increasingly deciding for themselves when to fire and what to fire at.

I believe that in wars of the future two major features will be asymmetrical war (what we call "insurgency" -blurring the boundaries between "civilians" and "soldiers") and increased participation by increasingly sophisticated autonomous robots/remote control weapons. (see this column for some thinking along these lines). I'll ignore the first aspect and focus on the second, in the interests of keeping the size of this post manageable.

Here is how robotics will develop in war. First, robots will engage in non lethal activities like mine clearing or I.E.D detection. (This is happening today) Then you'll see them accompany human combat units as augmenters and enablers on real battle fields. (This is beginning to happen) As robotics gets more and more sophisticated, they will take up potentially lethal but non combat operations like patrolling camp perimeters or no fly areas, and open fire only when "provoked" (This is beginning to happen too). The final stage will be when robotic weapons are an integral part of the battlefield, just like "normal", human controlled machines are today and make autonomous or near autonomous combat decisions. These robots will look nothing like the vaguely humanoid robots of Star Wars - they will have shapes suited to their roles - think tiny dragon fly / cockroach like systems to computerized tanks and helicopters.

And this type of robotic battle system will not be confined to the major superpowers. As of today, what gives robots an edge is not really dependent on hardware, but on sophisticated software, something that does not depend on superpower status. Someone in Korea or India or China, or even a semi military corporation can gain a temporary edge in robotics software comparatively easily (though the Red Queen Effect will kick in soon enough to close the gaps) sufficient to tip a battle or a war one way or another.

The funny thing is that quite a bit of the required technology is available, sometimes in embryonic form, today. "The future is here, just not widely distributed" - as William Gibson said. I know that the Indian defense forces have made substantial progress in battle field robotics systems (I can't be more specific - the details are classified so don't ask! :-D ), and what India is on the verge of doing, others are, too.

None of this is to claim that future combat will occur between machines, all clinical and detached with no blood being spilled. On the contrary.

Historically, every innovation in technology has been hailed as bringing in an era of bloodless warfare. When the airplane was invented, for example, many columns were written on how they made war obsolete because, obviously, with planes roaming the sky no army could make unobserved movements and so war was now futile! Dresden and the Blitz and Hiroshima were just around the corner.

My prediction is that war will become more lethal, with increasingly sophisticated autonomous weapons systems and ever more vicious asymmetrical warfare twisting the business of making "the other fellow die for his country" (or ideology or religion or ...) into newer and ever more horrific forms. And that is something that all of us working on robotics or machine learning should think about.

Update: First Armed Robots deployed in Iraq . While these are not really autonomous systems (they are remote controlled) and substantial technical challenges remain on the path to truly autonomous systems, people conducting research in robotics know how fast these barriers are falling. Those who are not researchers in robotics/ai may mull over the progress in autonomous navigation of vehicles from 2004 to 2007 as exemplified by the DARPA Grand Challenge. Now extend that rate of progress to areas other than navigation.

Watching the DARPA Car Race

It is almost 3 in the morning here in Bangalore as I type this. I've been watching the webcast of the DARPA Car Race. This years challenge was for the autonomous ( == no human control) robot cars to accomplish three sub-missions in an urban environment, following all traffic laws (obeying traffic signs, recognizing and conceding right of way etc). This task is a magnitude tougher than finishing a 130 mile autonomous drive in the desert, with zero human control, which was the goal of the DARPA 2005 car race.

Stanford's entry, Junior, just pulled in to the finishing area. While it isn't yet clear who won (because this is not a first past the post race), I'd put my money on Junior. (His (its? :-) ) predecessor Stanley won the 2005 Grand Challenge. This is history being made before our eyes.

The sight of the cars driving themselves, with the steering visible through the windows turning with no human hand at the controls is .. exhilarating... exalting... too cool for words. This is **REAL** programming in action. The people who made this happen are just incredible. Congratulations to all!

How lucky does one have to be to live in these times?

Now to drag myself off to bed (and dream of autonomous vehicles).

Snapshots from my Robot Simulator

I've built a tiny robot simulator for my ongoing research. Like the (absolutely brilliant) Stage Simulator, my simulator enables the simulation of robots in a 2D world. Of course my simulator is much simpler than Stage, primarily because it doesn't try to do as much as Stage does. On the other hand it has a couple of important differences that are important to my research. For e.g. it has a built in sensor and motion error model and (being implemented) a sophisticated Sonar Model.

Anyway, here are some early snapshots of a simple environment.

A simple world for a robot to navigate - The world is 10m X 20m with a resolution of 2 cm. The corridor is 2.5 m wide.

A robot (the little blue dot)stands in the middle of a corridor.

The world overlaid with a quadtree that overlays and recursively decomposes the 2D world and enables efficient ray projection.

The world illuminated with a laser. The illuminated points are rendered in yellow. The scan has a resolution of 1 degree,a "pan" of 360 degrees and a range of 3 m. I was a bit concerned that this would be slow but it takes no appreciable time, even when the angular resolution is 0.1 degree. All hail the quadtree!

Of course the robot sees only the illuminated points, not the complete length of the walls etc.

This is what the robot "sees". There are other sensors simulated (e.g. an odometer) but they are not rendered here. You'll notice that the laser sensor senses the environment "perfectly", i.e the range detected is the actual range to an obstacle that intercepts the laser beam. No real sensor works like that. To simulate this imperfection in a simulator, a probabilistic error model is added to filter the reading so the robot sees an approximation to the real range.

Here we've added a bivariate gaussian error (with a standard deviation of 10 cm) to the laser sensor. If you squint, you can see that the readings now only approximate the wall surface. To see this more clearly, we'll remove the unseen parts of the corridor walls and the overlaid quadtree to get a picture of what the robot sees, with the simple error model in place.

This is what the robot "sees" through the error filter. Note the "scatter" of the readings. Now the robot doesn't know for certain where the walls (or whatever objects obstruct the laser - different parts of the world react differently to different sensors - some surfaces absorb light for e.g returning no readings to the laser sensor) are.

The robot simulator is not the central thrust of my research effort, but it helps to have an environment in which one can plug in various sensor models, motion models and error models and see what the robot would see. I thoroughly enjoyed coding this up.

The next step is to replicate a realistic sonar (vs laser) sensor. There are a few interesting issues to be resolved to get this to work properly. Sonar works differently from lasers - cones of ultrasound have different propagation and reflection characteristics from rays of light. Stage simulates sonar with the same "ray trace" approach as is used for lasers. For my research I need a more accurate model (and hence the need for a custom simulator). Early next week, I will be visiting the Robotics Lab of the DRDO (The Defense Research and Development Organization - For non Indians reading this, they are the local equivalent of DARPA) to talk to some scientists working on similar problems. I have a basic sonar model working already, the key challenge of which was to layer different continuous probability distributions and sample data from them in reasonably efficient fashion, but to describe that would make an already long post longer, so I'll stop now.