Back to Main Page

Robots Everywhere!

RODNEY BROOKS - MIT Slides - 2.1 MB Download - 280 MB Watch the Talk (17:39)

Robotics is in the midst of a transformation. Robots are programmable physical machines that have sensors and actuators, and are given goals for what they should achieve in the world. The first deployed robots were in structured environments such as automobile assembly lines. At that time, computation and sensors were both very expensive, so the environments for robots were specially constructed so that robots could effectively operate with little sensing or computation. More recently, most research in robotics has been targeted at extending robot capabilities to unstructured environments - environments not prepared specially for them. The opportunities are enormous, as are the research and engineering challenges.

The US invented and led in manufacturing robotics, and subsequently was surpassed by Japan and the EU. The US currently has the lead in the new class of autonomous robots for unstructured environments, both in research and development. World demographics are pulling on robotics for unstructured environments, as demand for specialization often goes unfilled. Japan, Korea and the EU have made robotics for unstructured environments national priorities. Investment in research at the intersection of computation and robotics is critical for success.

The first industrial robots were built by a US company called Unimation. These robots didn't have much in the way of computers or sensors. They were expensive and they worked by being able to perform a motion, accurately and repeatedly, for years at a time. The dominant cost of developing these robots was systems integration since the computation required was not complex.


However, today we are seeing a new class of robots that no longer can be used by specialists only, but interact with ordinary people. Surgeons use teleoperational devices inserted through small openings in the skin during surgery, robots are used as autonomous cleaning devices in homes, and military robots are used by soldiers who have often never touched a robot before. Future combat systems, not yet deployed, include robots that would accompany every troop of nine soldiers. These robots, in development since 2002, rely heavily on computation, sensing and perceptual algorithms, user interfaces, reasoning under uncertainty, video compression, networks, etc. Almost every one is a US robot, and they are designed to operate in unstructured environments

In order for this new class of robots to succeed, current research must focus on increasing their autonomy. On one end of the autonomy spectrum is pure teleoperation, where every motion of every motor is specified by human operation. Supervisory operation of robots, where a human operator specifies motions and the robot decides the details of the operation, is followed by task-level autonomy, where a human operator specifies the task and the robot decides how the task will be accomplished. With total autonomy, we can send the robot off to choose its mission and determine the best way to accomplish it. One would expect that increased autonomy would bring a higher price-tag. However, current robot autonomy projects prove otherwise. The most expensive robot projects are almost purely teleoperational (e.g., the International Space Station), while the purely autonomous robots are usually the least expensive (e.g., the Sony Aibo). This occurs because, first, it is not worth a person's time to operate a cheap robot; and second, in must-succeed situations, it's too risky not to have a person overseeing the robot (e.g., the Mars Rover projects).

One concern critics raise is that there are too many people in the world to justify the development and widespread deployment of robots, especially autonomous ones. However, demographic shifts will create an increased demand for people to perform everyday tasks that involve interaction with the physical world. In addition, robots can fill a demand for labor in unstructured, harsh, and unstable conditions. In order to address this labor shortage, robots need to become more autonomous and succeed more often. Research must focus on ways to make autonomous robots more reliable and find ways to make them understand the world better.

Japan, Korea and the European Union have all declared their desire to become world leaders in this new wave of robotics. Japan and Korea have both been advancing their efforts to develop technologies required for the commercialization of robots in order to become a new world-leading industry. Korea's Ministry of Information and Communication predicts that robots will be in every South Korean's household between 2015 and 2020, and has grouped more than 30 companies as well as over 1,000 researchers under its wing. The European Union, under Framework 6 and 7, has many robotics programs, some overseeing the development of over 100 projects each. Recently, a US project called the "Urban Grand Challenge" (DARPA), brought together researchers to develop autonomous robots that could drive on roads for six hours without a collision.

Future research must address two challenges. First, full robotic autonomy requires better contextual understanding in generic visual object recognition, social contexts, spoken languages (and sounds), and new sensors. Second, in order to do more things, robots need the ability to manipulate objects, to network into physical and virtual ad-hoc groups, and to interact in teams.