(I'm joining this topic much later than originally intended, but after
spending a few hours on researching and composing a post, I stupidly
let it get eaten by a web browser. Please pardon the fact that I am
responding to the positions in several of the replies all at once.)
1) The Task From Outside the Box: Teleoperation Instead of Total
Automation
a) The purpose of a goal...
As MLW points out, the task is the thing that defines the difficulty.
Building a line-follower isn't a groundbreaking achievement. It's not
EASY, but neither is it likely to be costly.
MLW suggested a "mail delivery" robot. There are already robots that do
this. True, they aren't built by amateurs for under $1k, but it isn't
rocket science and IMNSHO it isn't particularly inspiring, either.
There isn't a comp.teleoperatedmachines.misc, so I think that tossing
teleoperated semi-autonomous devices into this box isn't too
unreasonable. You could get pretty-close-to-real-world-applicable
projects that are as starry-eyed as the X prize, but conceivably within
the reach of a high-school club.
b) the case against full automation
.i. the "Grand Challenge"
The DARPA "grand challenge" is a really stupid idea. The army doesn't
really want fully automated trucks. For one thing, if they ran into
combat, they probably would not be able to distinguish a camouflaged
soldier with a grenade from a bush.
They should also forget the idea of giving a set of coordinates to be
traversed ahead of time. Some of the contestants used a pre-processed
database of the terrain and, in the short period between being given
the coordinates and the "go," they intended to massage the routing and
obstacle avoidance information. In actual use when the supply vehicle
is supposed to rendezvous with troops on the move, the destination
would be likely to change, so the path would be likely to change as
well.
What DARPA should be asking for is a machine that can do MOST of the
course without guidance, but which can be teleoperated.
.ii. Man & Machine in Space
I'm against manned space exploration. I think that, until we get the
kinks out of making a 20+ year closed-cycle environment, we meat
products should stay in gravity wells. Once we are able to make roving
habitats that, once they go up, they aren't *supposed* to come down,
*then* we should restart manned space flight.
The rovers on Mars have accomplished amazing things, but I think they
could have done much more if they had been designed to allow for
complicated human interaction. They *did* use human interaction and
intervention, but the provision of improved communications should have
been done first (or at the same time) so that humans could help them do
more. AFAIK, Humans did not take the rovers out and play with them in
simulated Martian terrain, so the training level of the operators was
pretty low, anyhow.
.iii. machine < man < man & machine < man + machine
I've heard the term "sliding autonomy" used for some teleoperated
machines: they're autonomous to some extent, but they are run by a
human via remote control. A concrete example might be a 'twebber' (TWo
whEel Balancing Bot) that maintains its own balance and controls its
own motors, but which is directed remotely. Another might be a robotic
submarine camera which must be able to maintain position, attitude,
orientation, and ballast on its own, but which goes to goal
destinations provided to the operator.
.iv. if I were in charge...
Some contests that I would sponsor if I were wealthy would be for
"tele-explorers." A extraterrestrial rover that operates with human
interaction with a baud rate of, say, 32kbps and a one-way transmission
lag of 200 seconds or so, would be a great example. An Antarctic
explorer that did the same thing with 56kbps and a 3-second one-way lag
would be another. A submarine that operated with no lag but a 4.8kbps
transmission tunnel is another. Use your imagination and you can
probably come up with at least four more that are within general
technical reach and which would have a "Wow!" factor of greater than "a
mailman" (no slur is intended against the Post Office, especially those
employees who stock automated weaponry).
The instrumentation that would be needed for these devices is obviously
completely out of range for a "budget limited" project. There will be
no spectroscopes, diamond drills, etc., on an amateur budget,
especially of your typical high-school student. The "get it there and
so it can work" part, though, is not.
Any special equipment could be simulated. An XT-rover might be required
to collect a specimen of soil from three different kinds of
places--perhaps marked with large orange targets--and put them into the
"analysis tank" (a cylinder of specified size, perhaps a coffee can).
One could argue that he device could be little more than a "remote
control toy". I could reply that just building this part is already an
"amateur hard" project, but that's not the big reason that this contest
is "aiming high:" the part that would make this contest different is
the lag and bandwidth limitation. The "robot" would be placed in an
opaque tank (i.e., a box) and communicate (by cable simulating RF?)
only through whatever the student had built into it, using only those
sensors on board.
Sensors are what make for a useful device. A robot that contains none
could be called autonomous, but it would not be what I am talking about
when I say "autonomous robot." Creative use of sensors--which almost by
definition means "human" at the current state of AI--is what makes a
useful device more useful.
To me, that sounds like a really terrific contest with lots of
potential for real research--even by high schoolers--to occur. That's
what the X prize was about. It wasn't just to prove that you could get
to "space" without a government paying the bills. It was also to put
together something that actually might be able to be used again, and
that might actually make civilian space travel a reality. Exploration
of the Earth, for all its successes, still has left over 80% of the
planet unexplored (especially if you count the part below the crust;
unless you are in orbit, there's a place 15 miles from you no human has
ever seen: straight down). This sort of technology has the potential to
allow us to do things that we can't even begin to imagine.
2) The Journal
Part of reproducibility is to keep a notebook. I feel that a journal of
some kind describing the parts, techniques, failures and equipment
should be a big part of the scoring...perhaps as much as 90%(!). (A
blog falls into the category of "journal" in this context.)
Once you "handicap" for resources and score for the journal, it should
be possible for a robot to win that performs very poorly on the task
itself or breaks during the contest. A teleoperated machine that was
constructed for $84 using three hacked remote-control toys and a
soldering iron should beat a gizmo that cost $2300 and was built at a
machine shop, even if the latter were first place in
time-to-complete-task, and the former only succeeded one out of three
tries.
First off, research isn't research unless it's written down. You can't
reproduce something if you don't know how it was produced in the first
place. Process documentation is a good habit to inculcate.
Secondly, the "levelness of the field" regarding resources and
resourcefulness can be evaluated more thoroughly. A notebook could
serve to document total amout spent plus "free" donated (or fabricated)
items, services and equipment. (See #3 "Level Fields," below)
Third, it enhances the "contest" feel of it because people can see the
blood, sweat and tears in black and white. If you watch a plastic box
on wheels deliver mail, there isn't much excitement there. If you can
read about it in detail, though, it makes the participants much more
vivid in the mind of the observers.
Two measures of success in a contest are: "does it inspire people to do
more and better?" and "does it make someone else thing, 'Wow. That was
cool. Maybe I should sponsor a contest!'" With a notebook or journal,
this could change a robot contest from the excitement level of watching
a chess match to something on the order of a tale of inventiveness,
perserverence, and challenges overcome.
3) Level Fields
I think setting a dollar limit is unrealistic. For one thing, if I go
to the junkyard and pry a hydraulic system from a car and pay $5 for
it, what's the value of that? If that works, and next year someone goes
and pays $180 for it "off the shelf," how would you know it wasn't
salvaged equipment? If a guy talks someone into fabricating a $6 piece
of ABS into a part for him using a CNC mill, should he be punished for
access to superior equipment, even if he generated the CAM files
himself?
I agree that the "external" resources, both physical (e.g.,
oscilloscope, CNC mill, a fat wallet) and mental (e.g., dad's a
mechanical engineer, mom's an electrician, and sister Sue has a welding
company; close proximity to the library of a major University) make for
an unlevel playing field. That's life.
I agree that people who "make do" with less should get more credit.
There should be guidelines as to what is considered "more," and these
should be *subjectively* evaluated based on the journal. There is no
objective rule that would really give what I think those who have
expressed opinions about the "more with less" they want that would also
be practical and meet the "fair and reasonable" criterion.
Just going by "reproducibility" isn't totally fair. If I get out my
knife and whittle a doohickey, then write down 2"x4"x8' pine board: $5,
it is unlikely that you could reproduce my doohickey for that price.
Plus, I'm using my neighbor's lathe to get it in roughly the right
shape. How much does that count? My dad drew up a prototype drawing of
the doohickey just before he went to Sweden to pick up his Nobel in
Chemistry. By the way, the doohickey is coated in an electrically
controlled adhesive that he cooked up for me. Do you count off for
that?
The idea mentioned regarding "anyone can buy anyone else's motor for a
fixed price" was probably intended as a philosophical approach than as
a real suggestion, but let me set it to rest just in case: we aren't
talking about a contest that is based on one specific skill. We're
talking about design + engineering + crafting + programming +
operational training (for the teleoperated version, anyway). Apples
aren't orangutangs, so an exchange will never be equitable.
The goal is to measure the *team's* ability to do a lot with a little;
to have something fun and exciting; to produce new knowledge in
robotics, both for the participants and the world at large; and to
reward a behavior of inventiveness, perseverance and productive labor.
Criteria should be set out so that people can prioritize their effort
accordingly, but I do not believe there is a way to make it completely
objective and/or rigid. A journal would allow for the subjective
evaluation to be openly observed, and "openly observed" isn't too far
from "objective."
Perhaps the judges could narrow it down to three and the voting could
be done with a 1-900 number. Let vox populi make the final
determination based on the journals (or journal highlights).