From aaai97-owner@cc.gatech.edu Wed Nov 6 12:32:11 1996
Return-Path: aaai97-owner@cc.gatech.edu
Date: Wed, 6 Nov 1996 10:35:15 -0500 (EST)
From: arkin@cc.gatech.edu (Ronald Arkin)
Message-Id: <199611061535.KAA12905@arkin.cc.gatech.edu>
To: aaai97@cc.gatech.edu
Subject: Summarized responses
Cc: arkin@cc.gatech.edu, bonnie@cs.upenn.edu, hamilton@aaai.org,
kuipers@cs.utexas.edu
Sender: owner-aaai97@cc.gatech.edu
Precedence: bulk
Status: RO

Fellow Competitors,

What follows is a summary of the responses to the proposed events.
( The contributor names have been removed.)
Closer adherence to the rules in the future would be
appreciated (I asked for one sentence each :-).
Feel free to continue the discussion either in person with
Jim or myself this weekend at the AAAI Fall Symposium or
also on our web page (under discussions)

http://spbtrc.gtri.gatech.edu/AAAI97

We should have some final decisions to report in about a week.

- Ron
==============================================================================

Summarized Responses to Events

==============================================================================

Event 1 - Find Life on Mars
-------------------------------------

***

I like it, may compete.

***

Would I compete: well my entry is kind of slow, so if you don't mind, I'm
going to start collecting "Mars rocks" on July 4. I guess the conference
arena won't be open yet, so I'll probably have to use another venue...

Seems like a great contest. My only quibble is that bringing rocks to the
lander is probably as important as Mars-life -- though recording Mars life
would be the most important of all. However, squiggle balls are more fun
than rocks, so I'd leave things as you have them.

***

A. I like this event (see B), but I'd prefer if there were opportunities
to use novel sensing strategies to distinguish the objects (i.e., make
them "warm", or emitting some "gas").
B. Since I consider it likely that my team would only be able to field
small robots (probably even smaller than the AAAI-94 treaded vehicles),
this event may be our best chance to compete (i.e., it seems do-able to
some extent with a multi-agent robot team).
C. Not applicable (I would hope to compete).

***

I hope there will be no severe restrictions on robot size and weight (its a NASA
mission, but they don't share their budget with me), and terrain will be mostly
flat floor, so conventional wheeled robots can cope with it.
I'm planning to compete in this event - at least in "map where the objects are
located" class.

***

1. Life on Mars: This is cute, would be fun to watch, but seems
like an ad-hoc problem that has little relevance to real-world
robotics applications.
Whether or not you think you would be willing to compete in each event.
Not likely for event #1, unless it has a more serious flavor
(we have to be serious here at the Lab :-| ).

***

As I read the "Find Life On Mars" event I got really excited about it.
I think it would be a great event!! So I'd just like to cast my vote
for it.

While the others (find the remote and vacuum) seem interesting, I
honestly the Mars event will prove most intriguing!!

***

Event 1 is probably more fun than event 2, but is not clearly described.
These two events would seem to require similar robotics capabilities (vision,
mapping, moving small objects), although it might be really hard for a robot
to pick up (trap) a squiggle ball (I don't know how fast they move), so that
may be an extra layer of complication. On the flip side, some of the objects
for event 2 could be mighty hard to pick up (e.g., a styrofoam cup is easily
squished), and maybe squiggle balls are less likely to get damaged.

Willingness to compete:
Not very likely. But I'm not entirely ruling it out either.

Rules Committee:
Sure, I'd be happy to help.

Longer comments on specific events:
(more than you asked for)

Event 1: Life on Mars

I'm confused about what's intended. In one place, you say you want the objects
retrieved to be sorted into three classes (moving, stationary-1,
stationary-2), and I'm not sure what's different about the two "stationary"
classes. But in another place, you say that the robot's goal is to collect
martian life forms (which, presumably, are examples of "moving"). From my
reading, you either want the robot to collect only the life forms, or you want
it to collect "everything", but sort it correctly. (And for the other class,
you just want "everything" mapped out, but not actually collected, but then I
don't know how a robot can map out the moving martians.)

***

This sounds like a good event, and we would like to participate, at least
in the mapping competition. How large are squiqqle balls and how quickly
can they move? If it is possible to scoop them up and place them in the
lander without the need for manipulation (i.e. just using a shovel of some
short on the ground), then we may also compete in the other part.

Question - what is meant by sorting into 3 classes of objects (moving,
stationery-1, and stationery-2)? These only refer to the martians, right?
How are we to distinguish among them? The rules also do not make it clear what
distinguishes martians from rocks, and how "droppings" could be detected.

***

Sorry to be responding Sunday instead of Friday. I am concerned that
All three proposed events depend on vision and manipulation. We are
adding an arm and have CCD cameras, but the contests seem to emphasize
those areas and have little dependencies on navigation, map building, and
planning.

I would also suggest that each contest might have divisions (single
robot versus multiple robots, simulated vacuuming versus actual pickup).

***

A. It seems that it is too similar to the last squiggle ball event. It is
obviously more complicated, but is it too similar and will bore the
competitors/audience that have 'done that, been there'?
B. Cybot does not currently have a camera.
C. Sure.

***

1. A) This looks like fun, especially if the "squiggle balls" are very
dynamic. It might be interesting (though dangerous?) to make this a
"race for life on Mars", where all competitors are in there at the same
time...
B) Yes, I'd like to compete
C) and I'd like to be on the rules committee if not
***

==============================================================================
Event 2 - Find the Remote
-------------------------------------

***

Another good contest. Assuming you do not allow tagging of the objects or
virtual manipulation, I expect this contest can go on for a number of
years. This is the one that will stress the AI in the robots. I don't
expect I would compete in this one this year.

***

A. This is definitely biased toward larger robots (tall enough to reach
shelves and tabletops) with advanced vision and an arm. That is not
necessarily a criticism -- it may be the intent.
B. The likelihood of having a capable robot (mainly in terms of size and
manipulator capability) is very small for us, so I would not expect to
compete.
C. I'd consider serving on the rules committee.

***

Again, the same concern from the other end of range - if you use human size
environment with six feet high shelves, it will be way to high to reach for my
robot.
I would like to compete in this event if manipulator for my robot is ready and
it can reach most of the places with competition objects or virtual reaching is
allowed.

***

2. Find remote: I like this event, because the "fetch-it" robot
is a popular concept with the Average Joe Citizen; I'm all for
generating robotics publicity that increases the interest of
the general public in robotics.
B. Whether or not you think you would be willing to compete in each event.
Probably not for event #2, although it sounds interesting.

***

Event 2: Where's the Remote?

I think you intend for the scoring to be based on the last of the n runs, but
that's not clear from the writeup. Also, you say that the scoring is the
number of objects returned to the judge, but that robots that don't reject any
items get an extra point for each item returned....so in effect they get
double points? I suspect the time penalty is sufficient punishment. A robot
that rejects an item should still have a chance at winning.

***

Event 2: Where's the remote?

We probably would not enter this event as it requires more
sophisticated sensing and processing than we will be capable of.
We certainly would not be able to compete if manipulation is required;
perhaps the robot could simply approach the item in question and then
return to judge?

General comment:
With the exception of the first event, there is no provision for competing
in a "simulated" manner, so that, for example, manipulation is required to
compete in event 2. This seems somewhat limiting for a mobile robot
competition, in terms of how many teams will be able to compete.

***

A. I'm afraid this is beyond the capabilities of many of the teams. It
requires a good manipulator (which many don't have) and a good camera (some
don't have). This might be better if there were not so many variables of
height and furniture. Remember, picking up objects at arbitrary heights and
orientations is a very difficult problem.
B. Again, no camera.
C. Sure.

***

2. A) There needs to be an option for short robots, and robots with
grippers designed for the floor only! How will the robot be "asked"?
B) None of our robots are capable of the task as specified. Our
vision system only sees color blobs...
C) I'd be interested in the rules committee

***

==============================================================================
Event 3 - Home Vacuum
-------------------------------------

***

I love vacuuming! I assume teams of robots are OK. I will try and put a
system together for this one.

***

A. This can be done by an extremely dumb robot with minimal sensors,
especially if you allow the use of a beacon back in the closet, but
(again) that may be the intent.
B. There would be a small chance that we could compete here, but it
probably would require that someone independently attack this problem,
not using the same robots we would tailor for event 1 (I picture
something dumber and big enough to carry at least a Dustbuster).
C. If I established early enough that we weren't going to compete, and
if I wasn't also involved with the rules committee for event 2, I would
consider serving on the rules committee for this event.

***

That's also my wife's favorite and subject of my current project - requirement
of discharging the dust by robot is very hard, it requires a non-trivial
mechanical invention, which vacuum cleaner industry couldn't do in their hundred
years long history. I think most people wouldn't mind to change vacuum cleaner
bag once a month in exchange for having the house cleaned automatically.
I would like to compete if automatic dust discharge is not mandatory.

***

3. Home vacuum: As with #2, this is another task that the general
public would like to see in a service robot, and would find useful;
although perhaps this event would be more boring to watch than
some of the other events.
B. Whether or not you think you would be willing to compete in each event.
Perhaps in event #3,
depending upon the time commitment.

***

Part of what I like about part 3 is that if it's framed right, it provides an
opportunity for small robots that is not clearly available in the other
events.

Event 3: Vacuuming (This is also my favorite event :-)

Looks great; my only concern is the "confetti". It's probably harder to vacuum
it off of a linoleum floor (than off of carpeting), so it might require really
good (big? heavy?) vacuums. I wonder if this event should be staged on
carpeting. This would be fun for smallish robots if they are kept in mind --
maybe there should be two categories. (e.g., Robots that travel under the
furniture vs. those that don't.)

Hmmm. Two more concerns:
1. Discharging the dirt seems to me unnecessary and an added level of
complication. This could be considered the owner's responsibility, not the
robot's.
2. The "high concentration of dirt" would require additional sensing
capabilities to deal with. I think this could be fun enough without this
requirement.

The task would probably be better framed as "how much dirt can you clean up in
N minutes?". Then robots that could somehow identify the high concentration of
dirt would get a bonus for having that capability.

***

Event 3: Home Vacuum

We have some interest in entering this event, but have some qualms about
the need to actually vacuum. A small robot with a small vacuum could not
hold all of the dirt, and a larger robot would have a lot of difficulty
getting under furniture. Attaching a vacuum also adds a lot of mundane
complexity, including things like managing the power cord, that seems
to be against the nature of previous competitions.
---
having had a robot with
a dustbuster for a while now, I'll tell you that they don't pick
up much of anything. What I'd suggest is something more like coverage
of area with a dustbuster or vacuum (i.e. no actual vacuuming). Just
as an aside, the $30k commercial robot vacuum is about $10k of robot
and abotu $20k of vacuum according to the guy who came to Dave Miller's
symposium a couple years back...
---
that's about the best one could do, but good luck doing it with
an expensive robot -- I don't know about you, but I think
I'd rather not run my one decent beast over a pile of confetti --
that stuff gets in the "drive train" and its a pain to clean out...
seriously, I was just pointing out that this may be a level of
complexity that has little to do with robotics -- getting our R2
to "vacuum" in the hallways of this building was relatively easy
(it ran around carrying a dustbuster and did quite well at aavoiding
obstacles and covering the floor space, even went under tables and
things being short) - but it did a lousy vacuuming job with the
dustbuster...
---
btw, I'd be happy to be on
rules committee for tasks 2 or 3 which I'm pretty sure we can't
attempt, and would be happy to give some inputs for 4 (reception)
as we had Spiro serving cookies at a recent dept event and learned
a lot about robots as serving aids...

***

A. OK. Could be fun.
B. Possibly.
C. Sure.

***

3. A) Hmmm. Not too exciting given the static environment and map.
B) Maybe... but likely only if it was the only event.
C) Not too much.

***

==============================================================================

Event 4 - Hors D'oeuvres anyone?
-------------------------------------

***

I especially like the food touch!

***

Still like the idea. I think the public judges should be given several
coins, but that's a quibble. I would love to have an entry in this, but
don't know if I will get that far. If I don't enter, I'd like to help out
in any other way I can.

***

A. I like the concept of interacting, as well as the idea of spreading
out the competition to a different "arena," both in place and time. (But
don't use batteries as the "currency," since some people would KEEP them.)
B. I hate to keep belaboring this, but our robots would be small. If
people didn't mind bending WAY over (and if they would be reasonably
careful about where they step), we might be able to adapt our event 1 robots
for this event, too.
C. Not applicable, as I would hope to compete.

***

I'm not planning to compete in this event - vision system of my robot is not
robust enough yet to handle so dynamic environment.

***

4. Hors d'oeuvres: This would be a lot of fun, as it brings out
the idea of robots as entertainment; it would make a good "exhibition"
event.
B. Whether or not you think you would be willing to compete in each event.
Unlikely for event 4.

***

Looks great. And great for publicity. :-)

I wonder if there will be problems with people's personal electronics (pagers,
etc.) This kind of stuff could wreak havoc with sonars.

***

This event is very cool, and we would definitely like to enter it!
I assume that all the robots would be there at once? If so, perhaps
each should have a beacon so that they may recognize each other.
There is some concern about the noise level at the reception - would
anubody be able to hear the robots speaking? This problem s exacerbated
if they are all in there at once.

***

A. Could be messy!
B. Cybot can definitely pour drinks for people!!
C. Sure.

***

4. A) This is the most fun sounding, though the specifications are
unclear. Can we allow ask the guests to provide a color signal if they
want an hors d'ouvres?
B) Definitely!
C) Definitely!

***

==============================================================================