Interview with Michael Behar for a story in WIRED on Tactical Mobile Robots

February 26, 2002

By Michael Behar

1. Ray, in your opinion, do you think robots will be an integral part of warfare in the future? What kinds of tasks will they handle? And what kind of timeline are we talking about? Five years? Ten years?

RAY: We’re already using robot planes for both reconnaissance and actual weapon launching. Fully autonomous land-based robotic weapons are probably ten years away, but we’re close to systems that combine robotic systems with human control via secure wireless communication.

Surgeons are already performing surgeries on humans from afar using virtual reality systems that combine full immersion visual virtual reality with tactile haptic control. In this way the surgeon has controllable visual scaling over size so that small features appear much larger (e.g., the eye can be the size of a basketball). The doctor can also feel the tension when pushing on a blood vessel or tendon. While the doctor is in this immersive environment (reflecting the conditions in the actual remote patient), she is controlling a robotic system to perform the actual surgery. Thus the robotic system is effectively a remote extension of the doctor.

Although fully autonomous robotic weapons will ultimately be used, the next generation of remote robotic weapons will work similarly to the remote surgical systems. Why put a human in a tank if a remote human over a secure communication line works just as well? And if you no longer need to put several 180 pound humans in the tank, then you can make the tank the size of a dog, or ultimately, that of an insect.

Although the high level intelligent decisions are made by the remote human(s), it is still useful for the tank to have lower levels of intelligence for such tasks as navigating rough terrain automatically, climbing stairs (without requiring the human to have to navigate the logistics of each step), and so on.

Over time, these remote systems will become more autonomous as they gain greater sophistication in higher level decision making.

The development project known as “smart dust” is developing insect sized flying robots that can perform surveillance functions and communicate findings back to satellite collection stations.

2. DARPA and the R&D labs its been funding for the past five years has produced a pretty sophisticated breed of something they call Tactical Mobile Robots. Have you had a chance to see any of these in action? Do you think the TMRs that are out there now could be used in battle or for reconnaissance in an actual conflict somewhere in the world? How would today’s TMRs be used? Or do you think the research still has a ways to go?

RAY: I’m familiar with the TMR program. I’m quite familiar with robot development at a number of research labs including MIT, and CMU. I have not been given a specific tour of TMR robots.

It is helpful to put DARPA research into perspective. DARPA has played a pivotal role in developing a number of key technologies that have ultimately had important military applications. However, that does not mean that the actual devices created through DARPA funded research are used directly in combat.

For example, DARPA funded basic research in speech recognition through the 1980s and 1990s. Today, derivatives of that research are used for automated word spotting of intercepted phone messages. Of course, DARPA created the Arpanet, the forerunner of the Internet.

I don’t expect the specific devices from the DARPA TMR program to necessarily be used in the next round of warfare, but they are creating some of the basic technologies that will be applied to future generations of practical weapons.

There are already systems that can use binocular and triangular vision systems to create reasonably accurate three-dimensional maps of obstacles.

As I stated above, I think the threshold we are on are collaborative systems of robot weapons with enough intelligence for basic navigation and mobility, with remote sensing control by humans. We’ll see quite a lot of fully autonomous weapons within a decade.

3. What are a couple of the technological hurdles that researchers face in trying to develop a robot that can “think on its feet” (so to speak) in the heat of battle?

RAY: If human commanders are able to make certain decisions, such as “all humans in this particular area are considered enemies that can be destroyed,” then the next generation of robotic systems could detect any moving entities (which would include humans, animals, cars, etc.) and destroy them. Ultimately, if we expect autonomous robotic weapons to demonstrate human levels of intelligence, we’ll have to wait until machines pass the Turing test around 2029. But there are obvious advantages to robotic systems in that they can be more daring than a human, have the inherent advantages of machine intelligence in terms of accuracy, memory, and speed.

4. Is there any advice you would give to TMR researchers? Are they headed down the right path with the current designs of TMRs? Or are there other/additional avenues of research that you believe tactical robots should be directed toward?

RAY: The research program includes basic research on such areas as machine vision, three-dimensional mapping, and automated systems for mobility. These will be useful building blocks for future generations of robots for both military, industrial, and even home applications.