HPM@S1-A@sri-unix (09/17/82)
From: Hans Moravec <HPM at S1-A> There are currently a couple of factories in Japan that run unattended part of the time, usually the night shift. In those, the lights are turned off, since the robots don't work by vision anyway. The easiest methods for automating most of the dirtmoving, processing, and construction would use short range radars for terrain and object tracking - or some other active system - and would probably work better without that huge radio noise generator in the sky. Don't sell robotics short, it is presently vigorously alive, and by the time the transportation is ready, the automatic machinery for unattended operation 90% of the time will be too.
REM@MIT-MC@sri-unix (09/17/82)
From: Robert Elton Maas <REM at MIT-MC> Re: Radar or laser ranging for robotics on moon. Indeed, if you can get automated robotics smart enough to map terrain, plot optimal routes to avoid danger and collect maximal moon-dirt for smelting, all in the dark (no visual help from Earth-humans for 2 weeks at a time), fine. My point about shutting down at (lunar-)night was when using TV cameras and 2.5-second-delay-human control during initial minimal-equipment setups for materials-bootstrapping and system debugging. Later we may use active (radar, laser) ranging to create 3-d models, using that instead of TV pictures, sending back to Earth not a TV picture but a 3-d image that is displayed using some 3-d device. The human would still run the remote manipulator manually with the 2.5 second delay, but the visual feedback would be from the 3-d image rather than directly from moon-based cameras. Perhaps we'd use direct TV in the day and active ranging at night. During day we'd do long range terrain mapping and largescale planning, because daylight is of constant brightness regardless of distance whereas laser ranging brightness falls off inverse-square on the return path even if the laser outpath doesn't diverge. (Both fall off inverse-square for resolution, although that can be corrected by simple magnification. Using a large lens to collect more light may compensate for inverse-square light falloff, using magnification to reduce field of view to reduce noise from Earthshine. Thus with careful design maybe laser ranging can be used at night over long distances too.) I think a human teleoperator with various forms of computer-enhanced optics for visual feedback will be workable long before true robot control will be, and then just like on Earth we'll find tasks that the robot can do well and assign them to the robot, leaving the diminishing-rest to the human.