To: aaai97@cc.gatech.edu
Subject: Clarifications for the "Life on Mars" Event
cc: reids@CS.cmu.edu, tch@omegainf.cs.utah.edu, doug@mobile-intelligence.com
From: Reid Simmons <reids+@CS.cmu.edu>
Date: Mon, 24 Mar 1997 11:45:10 -0500
Message-ID: <5605.859221910@jubilee.learning.cs.cmu.edu>
Sender: owner-aaai97@cc.gatech.edu
Precedence: bulk

Here are some answers to questions about the "Life on Mars" event.
Updated rules will be out shortly (within a day, or so).

> What will be the size and weight of objects to be picked up by robot?

> It appears that some of the rocks will be very close to the size of the
> Martians. How many rocks will be close to the size of the Martians, and will
> they be close in shape as well? If the rocks are that similar, than that
> requires a greater reliance on vision systems to distinguish color and
> alienate other methods of detecting the Martians. I might actually try and
> avoid using vision systems.

> Can the rocks be divided into two distinct classes? Class 1- big ones that
> are obstacles (visible with typical sonar detectors), Class 2- small
> ones that could be confused with life forms. Colliding with the
> small ones should not be penalized.

> Can we know in advance what color the rocks will be ? (black would be good!)

The "Martians" will come in three varieties: (1) Stationary spheres, about the
size and weight of tennis balls, (2) Stationary cubes, also about the size and
weight of tennis balls, (3) Moving spheres, exactly the size and weight of
Squiggle Balls. Objects will be different colors (3-5 colors, total). All
rocks will be bigger than any of the objects, and they will all be
rock-colored (currently, the plan is to bring in real rocks).

> I can see I'm going to have a slight problem with the various colors, my
> robot only sees in black and white. Are black and white stripes an option?

All the objects will be distinguishable by size and/or shape. With only b/w
vision, however, you may miss getting points based on picking up the first
object of a given shape and color. Also, there may be a problem in detecting
the "danger zones" on the floor. For that, we will endeavor to make the
danger zones be a color that can be distinguished from the arena floor using
only a b/w camera. We don't yet know the color of the arena floor, so we
cannot yet say what the "danger zone" color will be.

> Will the perimeter be oval or rectangular?

It will be either rectangular or hexagonal, about 1000 sq ft in area.

> What is the resolution of the map?

The map will be hand-drawn, and will not be totally accurate. Teams can enter
the map information into their robot system however they like.

> What will the limits be on using a radio link to direct your robot? I want
> to avoid the radio link as well.

No limits on using a radio link, although in the past participants have found
them to be unreliable.

> I think coming within 6 inches of a life form is sufficient for non-
> gripper robots, as long as they announce the event in some manner.

The updated rules will say "come within 6 inches, announce the object, and
remain stationary for 5 seconds". While this is a bit hard to judge for the
squiggle balls, we'll try to do our best to be fair. Also, judges and/or team
members will remove objects from the arena when the robot announces them, so
that the robot cannot repeatedly "pick up" the same object.

BTW, there will be a separately judged category for non-manipulator robots.

> I can't tell from the rules exactly what the "pen/lander" will look like.
> In particular, do you envision the robots being able to enter the
> pen? How wide is the door? 

The "pen" will be a square, 1-2 meters in length/width (either lexan or
cardboard walls). There will be two ~0.5 meter width openings in the walls --
one for putting stationary objects, one for the moving objects. Currently, we
are not considering putting any distinguishing marks on the pen or near the
openings, but teams can do so on their own, for a penalty.

> Will all squiggle balls be the same color? Will this color be distinct
> from the dead objects?

Yes and not necessarily.

> If our gripper can grab a squiggle ball, can we assume it is large enough
> to grab the other types of objects?

Yes

> But I have one concern...is a trial really only going to be 5-10
> minutes? I'm afraid with the inherent size/speed limitations of
> our Nomad 200, we won't be able to do too much in that small amount of
> time. Can we ease that up to AT LEAST 15 minutes? 

First, we actually expect the trials to be 10 minutes (5 minutes is a lower
bound -- just in case some of the faster robots clean up when we are running 
10 minute trials). Second, why does it matter how long the time is -- you are
competing against other teams with the same time limit. That said, if we find
teams are not doing anything interesting in the first 10 minutes, we may decide
to lengthen the trials.

Let me know if you have other questions.

Reid