Page 64 - Modern Robotics Building Versatile Macines
P. 64

44   Modern Robotics


            Robot Mules


            An important early advance in many civilizations was the use of
            beasts of burden such as horses, mules, oxen, and camels to carry
            goods. In the past few years, Raibert has been a key researcher in
            a project to create the mechanical equivalent of the mule. Called
            “BigDog,” a prototype walking robot offers the military the
            ability to increase the carrying capacity of infantry. The robot



              SOLVING PROBLEMS: ROBOTS AND ANIMATION


              In 1872, photographer Eadweard Muybridge (1830–1904) set up a
              series of cameras at a racetrack. Each camera was set up so that as a
              galloping horse passed, a cord would break and a picture would be
              taken. The result was the first photographic study that revealed the
              exact sequence in which the legs of the horse moved. If one takes
              enough such pictures in rapid succession the result is a “moving pic-
              ture”—a movie.
                While film can be used to “decompose” walking or running into
              discrete movements, traditional animation is the reverse process. It
              attempts to create smooth action from a series of discrete images.
              There have been many approaches to animation, ranging from car-
              toons with hand-painted frames (cels) to computer animation that uses
              algorithms to create smooth transitions from one defined “key frame”
              to the next.
                The ultimate problem for animators is how to turn discrete images
              into smooth, realistic motion. One approach commonly used is to cap-
              ture the motions of live human actors and incorporate them into the
              animation model. Unfortunately, the ability to integrate such captured
              motion smoothly is limited—fundamentally, because the modeler lacks
              real understanding of the mechanics and dynamics. Without knowing
              why things move the way they do, the ability to determine where they
              should move next is limited.
                One of Raibert’s key insights is that the same algorithms that
              enabled his robots to walk or run smoothly and realistically could also
   59   60   61   62   63   64   65   66   67   68   69