NTB >> NEWS >> WHO'S WHO AT NASA July 2005
	Vladimir Lumelsky Lead Technologist - Robotics, Information Systems Division, NASA’s Goddard Space Flight Center Greenbelt, MD A robot sensitive skin engineered by Vladimir Lumelsky has been identified as a key technology to develop at Goddard. The high-tech covering would enable robots to sense their environment and react to it, much like humans respond when something or someone touches their skin. Such a technology, referred to as a "High-Tech Skin," is seen as essential for carrying out NASA’s Vision for Space Exploration.

NTB: What is the “High-Tech Robot Skin?”

Vladimir Lumelsky: The robot sensitive skin can be described as a front-end or hardware-end of a larger conceptual framework that includes strategies and algorithms for sensor-based robot motion planning. Robot motion planning is the central problem in robotics; it is a wide and general area. It includes motion planning for the robot hand, such as when manipulating objects; motion planning for specific operations, such as welding and painting; and motion planning that involves the whole robot body, such as when moving around and trying to avoid collisions. We humans are involved in motion planning all the time – when fidgeting with a pencil, adjusting to a more comfortable position in a chair, or walking down the street. It is only recently that we turned to teaching our robots to do similar complex actions. Motion planning becomes especially important when we want our robots to be able to handle a natural, unstructured environment – as opposed to a strictly structured environment of a factory floor where the environment is designed, usually at a high cost, to compensate for the robot idiosyncrasies.

The picture becomes even more complex as we uncover cases where positions switch: it is not only that humans can teach robots to do motion planning – it turns out that sometimes robots are smarter than humans in doing some specific motion planning tasks. It is not that we want the robots to teach humans in such cases. Rather, it makes sense to shift the responsibility for those tasks from humans to robots. For example, human operators that control robot arm manipulators at NASA Shuttle or on the International Space Station, or robots in nuclear reactors, move those robots extremely slowly, in order to avoid damaging collisions with surrounding objects. The low speed allows the operators to think carefully before they do a tiny motion increment, thereby making sure problems do not occur. Today we know the underlying problem – humans are not very good at cognitive tasks that involve spatial reasoning, hence the slow speeds. We would like operator-controlled (as well as autonomous) robots to move much faster, to improve the overall’s task efficiency.

Robot arm manipulators can do some of these motion-planning tasks very easily, while working under the same conditions as humans. This is not because robots have better memory or can calculate faster – it is because they can “think” better about space using some motion planning strategies and algorithms. Can’t the human operators learn those same strategies? It turns out they cannot, due to those same limitations in the human spatial reasoning skills.

For a robot to realize the said strategies, it has to have a special type of sensing – a skin-like sensor that covers the whole body of the robot, the way the human skin covers our bodies. In other words, as potential collisions may occur at any point on the robot’s body, the skin should uncover those situations as they appear, and point out which points of the robot body are in danger. Mathematics can then be used to build algorithms that allow the robot to avoid collisions and move safely.

Human skin works on touch principle. Because today’s robot arm manipulators are built of hard materials and move fast, such tactile sensing is not good enough for robot motion planning. A better technological solution is a skin that senses at a distance (proximal sensing). A couple of generations of the sensitive skin that we have built so far, based on the theories and observations mentioned above, are based on proximal sensors – namely, active optical infrared sensors. Physically, each infrared sensor is a pair that includes an LED (light emitting diod) and an infrared detector. Each LED on the skin sends out a beam of infrared light. If the light reflects off of a nearby object that happens to be in front of this sensor pair, the reflected light comes back to the skin and is detected by the optical detector. When covering a human size industrial arm manipulator, the current skin version includes about 1,300 sensor pairs. All this sensing and data processing electronics is installed on a flexible substrate that can be wrapped around the robot body as needed. To produce a continuous motion, including real-time reaction to potential collisions, all sensor polling and data processing occurs within the normal sampling rate of the robot arm.

NTB: How will NASA utilize this development?

Lumelsky: Just about any application of robotics at NASA, specifically in the Exploration Program, will likely benefit from the use of the sensing skin. Note, for example, that no humans are allowed to share space with today’s industrial robot manipulators. It would be too dangerous for a human to enter the workspace of an operating robot. This is because today’s robots are not sensitive enough; they do not have sufficient body awareness, and hence cannot avoid dangerous collisions. The result is that the companies utilizing those robots spend upwards of four to five times the worth of the robot itself on a specialized environment in which to use it. Besides the money issue, NASA’s applications call for a completely different type of a robot – a robot that can deal with the environment as is. A robot covered with the sensitive skin will be able to behave adequately in the environment as it comes.

In particular, we want the robot to be able to help our astronauts. We would like to have, for example, a robot astronaut assistant, capable of handing the astronaut various tools, taking them back, and storing them. To realize such collaboration, the robot has to work next to the astronaut – which is exactly what is not allowed in today’s industrial robotics. We have to break this barrier and make it possible for humans to share space with robots. That is the role that the sensitive skin will play.

NTB: What needs to be accomplished in order to make the skin space-ready?

Lumelsky: In principle, the feasibility of the idea has been shown. A couple of generations of a lab-scale skin have been demonstrated. We know, however, that for NASA this is not enough. The skin has to be redesigned to utilize space qualified electronic components. The skin configuration will vary depending on the robot it is used with. Skin modularity and universality will hence become a serious issue.

In space, one second the sensor may face the bright sun light and the next it can be totally dark. Since our sensitive skin works with infrared light, its optical sensors may be more vulnerable in space. Temperature extremes are a concern. The issue of radiation must be addressed as well, as the exposed skin electronics will be vulnerable to it. These and other NASA specifics will have to be accounted for.

NTB: Are there commercial applications that exist for this technology?

Lumelsky: There are definitely commercial applications for the robot skin. For example, today’s robot arm manipulators – with the capabilities that we see in the automotive industry – could be used to assist the elderly in their homes. However, you cannot put such a robot in someone’s home, because it is too dangerous. The robot would not only hurt itself – more importantly it would hurt the person it was trying to assist. The skin would help to solve this problem. Skin-equipped robots also would be able to work in agriculture, undersea, or on our streets.

Visit http://gsfctechnology.gsfc.nasa.gov for more information.