We may think robotics is all about machines, but attending the IEEE Technologies for Practical Robot Applications (TEPRA) conference on April 22–23, 2013, reminded me that vision and sensing also are crucial. Robots need to "see" and understand their environment in order to work efficiently and safely, and that requires sensing and information processing systems.
Low-cost sensors are essential for consumer devices such as the household vacuums and cleaners developed by iRobot (Bedford, MA). A $20 laser radar would about $100 to the cost of a robot, said Stefan Guttman of iRobot, so the company developed an infrared (IR) navigation system called Northstar to guide its Mint floor cleaner. Three IR sensors mounted on the sides of a pyramid in the cleaner detect IR spots projected on the ceiling at pulse rates of 2 to 6 KHz by a small cube illuminator placed in the room. On-board electronics measure the signals and compute the cleaner's position to within 7 to 12 cm.
PrimeSense (Tel Aviv, Israel) is building upon the success of its coded-light gesture-recognition system, best known for its use in the Microsoft (Redmond, WA) Kinect. With more than 20 million Kinects sold, PrimeSense introduced a more compact motion sensor called Capri at the Consumer Electronics Show in January, 2013; only the size of a stick of gum, it can be used in mobile devices. At TEPRA, Ohad Shvueli of PrimeSense described other applications in robotics and medicine, such as gesture control to avoid touching controls in sterile environments.
A structured light sensor can guide unmanned aerial vehicles (UAVs) around complex urban environments, says Helen Greiner of CyPhy Works (Danvers, MA). CyPhy has developed a lightweight laser scanner that can make 10,000 readings/s and draws only 2 W for use in an 11 lb UAV carrying a 3 lb payload. Controlled by a pair of wires running through a microfilament, the system was designed for inspecting buildings and other areas where GPS us unavailable for navigation during urban combat operations.
There is plenty of competition in robotic sensing. Rethink Robotics (Boston, MA) uses ultrasonic sensors as well as machine vision to guide its $22,000 Baxter robot, designed to be easily trainable for repetitive industrial tasks, which it displayed at TEPRA. Cost is unquestionably an issue. But the overall message was loud and clear—robots need to sense and understand their environment, and optics are one way to provide the needed "eyes."