Robotic mounts need roll-cages

When I was based in the old Camden Jim Henson’s creature shop (currently in the process of becoming a swanky block of flats), working with Dave Housman, I learnt about why it would be useful for motors to feel pain.

The Big One was a puppet control system developed by the creature shop to control complex motorised animatronics in a simpler way. Before the new control system, those puppets that could not be controlled directly by the hand-up-arse method had many servo motors in them, each wired to a dial or lever. It would take a team of people, each controlling one or two motors, to coordinate a simple smile. What the Big One did was to make a controller that had many degrees of freedom – basically a sensor filled kermit the frog – that allowed a single performer to map the movement of their hand in the control glove to the motors in the puppet (this later evolved into full computer mapping from glove and joystick movement to emotions of the puppets both real and digital in the Henson Performance Control System).

The problem was, if you ever pulled your hand out of the controller at the end of a good take on set of filming, for example, and let the jaw of your controller drop away suddenly, the motors in the giant dragon face in front of you would faithfully replicate the movement. Puppets could rip themselves apart – the motors were generally strong enough to do the damage. Similarly when testing a control system, if it was powered down and back on, it was possible – depending on how carefully the electronics and code had been thought out – for a motor to reset to an extreme position. Again this can be beyond the reach of the puppet, leading to broken jaws or limbs impaling themselves.

I am reminded, nervously, of these reset conditions when watching this video of an extreme thrill ride test using an industrial robotic arm. I know the developers probably ran the arm many times before climbing in, and that the accuracy of these industrial systems must be extremely high. But unless you put sensors in, like the doors on a lift, machines just don’t stop.

One answer is to give robotic limbs pain. Embedding sensors into joints that detect strain and limit the power of their motors, just as our own bodies prevent us from breaking ourselves most of the time. A second solution is to borrow from another machine that can have dramatic faliure modes – the sports car. A roll-cage is a structural defense – something strong enough to withstand your system going wrong. I think if I were testing an industrial robotic mount, I would want a roll-bar.

Leave a Reply

Your email address will not be published.