MINSKY@AI.AI.MIT.EDU (Marvin Minsky) (03/11/89)
I agree with Paul Dietz about this: I don't think this proves anything, except that a 5-second delay means you cannot operate a Go-Kart at normal speed. Need I remind you that teleoperation of a lunar rover was accomplished years ago by the Soviets? I met one of the Lunakhod drivers many years ago and he said it was fun to drive. Why are people so skeptical of telepresence when successful control of systems involving delay are all around us? I think it is because of the myth that the mind makes direct contact with the world through the body. Bad metaphysics makes bad engineering. Instead, one should consider a more realistic model of how we interact with the world: The human sensory-processing-motor loop takes about T= 1/6 second. Therefore, with delay D, we can work at speed T/(D+T). So, with 1 second delay, you should be able to work at 1/7 real time. For an orbiting space station, with good communication, the delay could be held to 2/3 second using geo relays, or to 1/6 second using a chain of earth-based or LEO relays - so we could operate between 1/2 and 1/5 real time speeds. There might be some special difficulties at the 1/2 speed rate. But I have seen no evidence that there are difficulties at slower rates. Such delays should be very tolerable, because the power and weight requirements for a telerobot should be, I estimate, over 200 times smaller than for a human. If we also recognize that a person can work attentively less than 1/4 time (6 hours/day), we have a payload gain of over 800. So, even with a slowdown of order 8, telepresence gains us a productivity advantage of 100 per unit mass in orbit! Accordingly, I believe that a telepresence-based space laboratory could do the same or better at a much lower cost. "Remotely-manned" is better than either "manned" or "automated". The telepresence equipment could surely be developed in 4 or 5 years, because the engineering is not especially hard. NASA should have done it already, but it is never too late to start. P.S I propose the verb to "teep", for operating things by remote control. Teeping is fun and safe. Marvin Minsky
ruffwork@mist.cs.orst.edu (Ritchey Ruff) (03/13/89)
In article <553934.890311.MINSKY@AI.AI.MIT.EDU> MINSKY@AI.AI.MIT.EDU (Marvin Minsky) writes: >[...] NASA should have done >it already, but it is never too late to start. Well, some people at NASA Ames Human Factors are working on the hardware for this. Mike McGreevy and Scott Fisher have been working on a "virtual workstation environment". It includes 3D head mounted display with motion and positional sensing, gloves that allow the computer to track the hands (it's "shape"---finger positions---and its location) and they are working on sound (so that as you walked around in the virtual environment any sounds would seem to stay stationary). They are also working in cooperation with JPL on a remote controlled hand: you move your hand in the glove, and the robotic hand mimics your action. There was a Scientific American (10-87) that talks of this work. They list possible applications like: - ad hoc repair and/or retrieval of GEO-sync satellites, - remote exploration of planets from orbit, - supervising automaticed robots. >P.S I propose the verb to "teep", for operating things by remote control. > Teeping is fun and safe. > Marvin Minsky I always liked "waldo" (trivia: which sci-fi author came up with this term?). --Ritchey Ruff ruffwork@cs.orst.edu
joe@hanauma.stanford.edu (Joe Dellinger) (03/13/89)
I remember reading an article describing how they train pilots of Oil Tankers. They sit them in a very small very slow motor boat in a small pond, and put HUGE delays on all the controls, like 30 seconds. The article said that for a while the pilots would crash the boat against the walls, etc, etc, but with a little practice they would learn to pilot it exactly where they wanted to go without thinking. Your body gets by with the "huge" delays in our bizarre electrical - chemical circuitry because the feedback loops can be cut short in certain critical cases. If you start to burn your hand, it jumps away from the heat literally before you have time to think about it. You can, on the other hand, also override such reflexes by conscious effort. Other more complicated things, like learning to shut your eye when you hear a certain tone, can be "hardwired" into your lower brain with a little practice. (Really!) Etc. Mammalian nerves carry signals faster than "more primitive" life's did, and yet 100 foot long dinosaurs whose nervous system probably took half a second to carry a signal from their hind feet to their head and back evidently walked around on irregular terrain at respectable speeds without tripping over their own feet. It seems to me that telepresence shouldn't be ruled out. We just need to train people to get used to it, and learn enough robotics to have some "reflexes" handled locally. \ /\ /\ /\/\/\/\/\/\/\.-.-.-.-.......___________ \ / \ / \ /Dept of Geophysics, Stanford University \/\/\.-.-....___ \/ \/ \/Joe Dellinger joe@hanauma.stanford.edu apple!hanauma!joe\/\.-._
ELIOT@cs.umass.edu (03/18/89)
A very simple experiment might provide some of the "touch and feel"
of teleoperation for those without access to a mechanism with
adjustable delays in the controls.
Try doing various things while wearing heavy gloves with a slow
strobe light for illumination. Better still, get one of the mechanical
claws that are used to remove merchandise from the top shelf of
department stores.
The strobe light causes a variable delay between the current state
of the world, and your information about it. This differs from
a constant delay, but should have some resemblance.
Chris Eliot
University of Massachusetts at Amherst