From: Nick Szabo Subject: NASA technology choices Date: 28 Feb 91 02:59:15 GMT Organization: Sequent Computer Systems, Inc In article <1991Feb26.192615.19617@zoo.toronto.edu> kcarroll@zoo.toronto.edu (Kieran A. Carroll) writes: [a very informative article on teleoperation, with which I only have a few nits to pick :-] >For the type of teleoperation that was planned for FTS (force-reflecting >master-slave set-up), it has been shown that closed-loop time-delay of greater >than about 10 milliseconds cannot be tolerated. Research done several years >ago at JPL, in support of FTS design, showed that the closed-loop system >starts to go unstable for larger time delays. * FTS was funded as part of Fred, was it not? Would a study funded with the _goal_ of operating FTS from a space station tell us what, in fact, is the most _economical_ way to perform FTS's tasks? * "tolerated" is a rather fuzzy word. Exactly what loss of functionality is generated by going over 10 milliseconds? * Is the "force-reflecting master-slave setup" in fact the most economical technology for performing FTS's tasks? >I imagine that this could be >ameliorated by increasing the time-constant of the joint controllers >of the robot, but this would make the manipulators more "sluggish", >which causes operator fatigue to increase significantly. Does operator fatigue cost even a significant fraction of the space station astronaut time required for on-site teleoperation? Training several operators for working short shifts on Earth would seem more economical by several orders of magnitude: labor cost on Earth: $30/hour (the best video-game players on the planet :-) WAG for labor cost on Fred: $30e9/(4*8*5*50*20) = $187,500/hour Even assuming 2 earth shifts for every space shift due to fatigue, we get costs on Fred over 3,000 times greater than costs of teleoperating from Earth. >...[Shuttle arm] The feedback loop is closed not by force-reflection, >but by the operator observing the motion of the end-effector (or payload), >either directly or via TV cameras. > >This method has the advantage of allowing much longer closed-loop delay times >without instability. While FTS could not be operated from the ground >(where the closed-loop delay could approach 250 milliseconds), a properly- >designed commanded-rate controller with TV-camera feedback could >do the job (as long as a reliable communications link existed). What tasks require FTS to have force-feedback? Can these tasks be dropped and still have a signficant subset of FTS tasks economically accomplished from Earth via video feedback? >Work >in Canada is proceeding to generalize this concept in ways that would >allow it to be used to perform Earth-based teleoperation of devices on the >Moon; one step in accomplishing this is to build some autonomy into the >lower-level control loops, with periodic (time-delayed) "supervision" >coming from the Earth. 3 seconds is quite long in teleoperation terms. Even more "semi- teleoperation", what I call "long-RTLT teleoperation", is being designed for the Mars rover. It relies on much more autonomy than a 3 second or millisecond teleperated device, but there are still ways to generate feedback (including mapping of future terrain and ground "pre-teleoperation" to perfect moves that are then uploaded to the rover). -- Nick Szabo szabo@sequent.com "What are the _facts_, and to how many decimal places?" -- RAH