Increasingly, robots are being used in environments inhospitable to humans such as the deep ocean, inside nuclear reactors, and in deep space. Such robots are controlled by remote links to human operators who may be close by or thousands of miles away. The techniques used to control these robots is the subject of this book. The author begins with a basic introduction to robot control and then considers the important problems to be overcome: delays or noisy control lines, feedback and response information, and predictive displays. Readers are assumed to have a basic understanding of robotics though this may be their first exposure to the subject of telerobotics. Professional engineers and roboticists will find this an invaluable introduction to this subject.
Imagine that you are sitting in a chair, perhaps even one with a view of the sea, and that I place a baseball-sized sphere in your hand, connected to the back of the sphere is a mechanical linkage that disappears up into the ceiling. Now imagine that on the floor at your feet I place a robot arm. On the end of the robot arm is another baseball-sized sphere and somehow, as if through magic, the sphere in your hand is connected to the one on the end of the robot. Now, whenever you move your hand left, the robot moves left. Whenever you move your hand up, the robot moves up. This mode of operation, where you, an operator, directly control a robot is termed teleoperation, and its historical development will be described in Chapter 3.
Now, imagine that you move your hand down, causing the robot to move down. When the sphere on the end of the robot contacts the floor, you feel the sphere in your hand stop moving. Even though your hand is still above your lap, it feels as though you had reached down and touched the floor yourself. Any force you apply to the sphere in your hand is duplicated by the sphere on the robot, and any force felt by the robot is duplicated on your hand. This is called bilateral teleoperation.
Now, imagine that I take the robot and move it to the other side of town. Then I place a TV screen at your feet and connect it to a camera pointed at the robot. Now, you can control the robot just as before. When you move your hand, the real robot and its TV image both move. Only now your view is a little more restricted; since the TV picture is flat you can't judge depths quite so well, and if you lean forward in your chair you just see the back of the TV set, and not the back of the robot as you could before.
Now, imagine that I take the robot and the camera and move them further away. You can still see the robot on the TV. It looks slightly larger than before, though perhaps that is just your imagination, and the ground around it looks strange. Curious about what it feels like, you move to touch it. But nothing happens, so you move some more, but still nothing happens. Then suddenly, after several seconds, you see the TV image of the remote robot begin to move, and you recognize that it is doing what you did several seconds ago. Pausing to think, you realize that the connection between your sphere and its sphere is not magic at all, for if it were, then things would not become delayed as they became further removed. But your musing is interrupted, for suddenly your hand is jolted upwards. Looking at the TV, you see that the remote robot has smashed its sphere against the ground---the force you just felt was the force it felt several seconds ago. Conventional teleoperation does not work well in the presence of communication delays.
To avoid problems caused by the delay, I could replace the TV with a computer-generated display and show you, not what the robot was doing now, but instead what it would do when it tried to duplicate your motions. When you move your sphere, the simulated robot on the computer screen responds immediately. You can control it (and hence indirectly control the real remote robot) in much the same way as when the robot was right at your feet. This type of system is called a predictive display. On its own, the predictive display is not sufficient, since the computer can't simulate the remote environment perfectly. Thus, to make the system work (and avoid smashing any more spheres) we need to add some local intelligence to the remote robot. Not only does that intelligence help protect the robot, but it also allows us to communicate with it using higher-level symbolic commands, thus making it feasible to communicate over links that have low bandwidth as well as high latency...
