The intertwined layers of electronic feedback and human spontaneity
Most of the electromechanical devices I am fond of playing with are not robotic per se but rather are simply remote controlled machines. However, there are similarities in that mechanical devices that are remotely controlled may have everything necessary to be proper robots except for the fact that they have no ‘brain’ built into them. In fact, it may even be best to test a potential robotic platform by making it remote controlled first, and then when all the bugs are worked out, one can move ahead with building in a brain and not worry about the platform.
Many of the so called ‘robots’ I have built as props for commercials and such are what is known in the industry as “animatronic.” The control systems on these are, for the most part, ready to be interfaced with a computer—the robot’s brain. There are a couple of reasons that we do not do this from the beginning. As mentioned above, we would likely run an animatronic puppet through its paces using remote control anyway, just to troubleshoot it. Secondly, there is a live-performance aspect to employing animatronic robots that is necessary and valuable on a movie or commercial set.
Directors will always want to be able to ask for immediate changes in a performance. Just like actors, puppeteers will often make a valuable contribution to a composition with spontaneous quirks or responses that aren’t scripted but which may end up being more important than what was originally in the script.
There are a lot of examples of what happens when live performance is taken out of a movie—think effectsladen films that are loaded with computer graphics. There is a well known science-fiction series (that shall remain nameless here) that was great when it was done with more hands-on effects, and lousy when it started to be primarily computergraphic driven. Not that one can’t do a good job making a movie purely with computer graphics. It just needs to be noted that creativity in a performance setting is important, and its absence needs to be compensated for if a director is determined to do everything with a computer. On the other hand, there are times when the artificial intelligence of a computer is critically important.
That brings us back to our original question: When should one drive a robotic device with a computer? On a movie set, the main place we use a proper robotic device is for camera control, or so-called motion control. This involves having everything possible controlled by a computer so that moves are exactly repeatable. This is an important part of the process of compositing various special effects. If for example, you have a miniature set and you intend for an actor to look like he or she is in it but want to have a camera move, you will have to have the camera move while the actor is acting in front of a blue screen, and then have the camera do a scaled-down version of exactly the same move while filming the miniature. Then you put the two together and they match. Having a computer robotically control the camera is the only way to go.
Another issue to consider is the nature of the thing you are trying to control. There is an interesting story about a problem I ran into with a device I had patented some years ago. The patent involved using a gyroscope to control a rolling sphere. Usually gyros are used to provide positional feedback and not brute force. While gyros can tend to stabilize or dampen movement, unless the rotational speed is very high or the weight of the flywheel used is very heavy, they will easily be forced into what is called “gyroscopic precession.” This is a tendency for the gyro to rotate on an axis at right angles to a force applied to it. That is, instead of keeping its rotational axis perfectly still and aligned and simply moving in the direction of the force that is being applied to it, the gyro moves off to the side.
My patent was for a rolling object like a sphere which has interior components concentrated on the bottom of the sphere. The motor is situated such that it is driving the weight of the sphere’s components forward; and because the exterior is round, it will roll the whole sphere forward instead of just swinging the weight around inside.
Now, these have been around for a while as toys; what hasn’t been around is that I used a gyro to dampen that weight inside. Without it, the ball would tend to have a pendulum effect, and every time you started and stopped, it would wobble back and forth for a bit making control difficult. I also have in the patent a feature that if the axis of the gyro is vertical, I could make the device turn right or left by accelerating or decelerating the gyro; the mass of the gyro flywheel resists a change of velocity with a sideways motion, so the sphere turns instead.
The problem with the device is that unless the gyro motor is really powerful and/or the flywheel is very heavy, you will end up sending the gyro into precession if you try to make the ball go up a steep hill or accelerate rapidly. If you do this, the ball will do a quick U-turn because of precession, which is frustrating. The only time I have used the device was for a Chef Boyardee commercial in which a can in the supermarket jumps off a shelf and follows a little girl home. I built the gyro control system into a plastic tube dressed up to look like the can, and we used it with some limited success on the commercial. Not only did the device have problems with precession when it ran into an obstacle or had to go up hill, but it was also really spongy in steering. It would be kind of like trying to steer a radio controlled car with a speed controller rather than with a servo, only spongier, or like driving a car by playing shuffle board!
Some years later I ran into a device that someone had come up with for the military as a reconnaissance device, which is my point in mentioning my invention in connection with control systems. The device used the same trick I did for controlling a sphere, but it had a GPS system and/or a compass which it would check at a certain frequency. Instead of a continuous acceleration on the flywheel, it would give it a pulse of acceleration and check, give it another pulse and check and so on, ratcheting itself into position. In this way—with computer moderation—it was able to take what was so spongy that it was hard for even a human to control and make it into a semi-autonomous, precisely controlled machine.
So should it be called a robot or a remote controlled device? In some cases, to do the job, you need help from control systems to the point that it is in fact what we would call a robot under a purist’s definition (which assumes some measure of onboard intelligence that permits a degree of autonomous action). You can make a case that we should include in this definition a Waldo control, or master/slave controls, and even hobby radio controls that have programmable transmitters with mixing and graphs to tune acceleration curves. Somebody, somewhere, had to figure this out and put together a machine capable of combining sequences of activities required to perform a given task. The robots that have been selling by the hundreds of thousands for sweeping your house do only moderately more than this, and currently the military is using very sophisticated remote controlled land and air-based devices that really blur the line in this area. The so-called robots that are either flying over hostile territories or the tracked robots that are investigating the same areas on the ground are remote controlled (the military is thankfully not yet sending out completely autonomous robots with weapons on them that I am aware of!), yet they have a number of features on them that are self-monitoring and automatic. These are legitimately presented as robots, but the reality is that it will be a long time, if ever, before something like that is completely autonomous.
What can be called a “robot,” in my opinion, all just gets down to a mechanical device with some sort of positional feedback mechanism, whether it is a simple servo in a radio controlled car, or a device that walks and talks like a human. But we probably won’t really be comfortable with calling a robot a robot until it not only blurs the line between remote control and autonomy, but also blurs the line between a machine and a human.
Jamie Hyneman is cohost of the popular MythBusters TV show on the Discovery Channel and is founder of San Francisco-based M5 Industries (www.m5industries.com), which specializes in animatronics and visual effects for commercials, movies, websites and print media. —the editors
Words by Jamie Hyneman