|
This page details the human-robotic interaction analog task for the CMU
MHCI group for Spring/Summer 2005. We describe the analog
environment, and the structure of the game including details about goals and scoring.
This paper does not cover robot assembly instructions, hardware setup
details, or software setup.
Environment Setup
The analog environment consists of a surface for the rover to drive on and a
set of targets arranged in accordance with the rules of the game. Although there is some flexibility in the exact
dimensions of the surface, it should be approximately 6 on its smallest side to allow the rover to navigate
between the targets. The orbital camera should be
configured to image the entire surface.
Creating Science Targets
Create between 12 and 16 science targets. Each target should be a four-sided prism (not necessarily a cube). Label the tops of the targets
with red, blue, or green dots representing values from 1-3 (approximately 1/3
of the targets should be 1's, 1/3 2's, and 1/3 3's). On the four sides of the targets, place multiplier labels for 1x, 2x, 4x, and 8x. Each target should have
exactly one of each multiplier value.
Target Placement
Place between 12 and 16 targets in a grid-like layout. Leave some
spaces in the grid and leave the edges ragged. Orient the targets so that the
sides facing adjacent targets have the same multiplier value.
Objective
The objective of the robotic task is to obtain the highest possible science
score in the time and resources allotted. The score is assigned based on the
number of science targets visited and imaged by the science camera.
Game Play
Users will have [fifteen] downlink opportunities and as many uplink opportunities as they wish. Each downlink will be separated by [two] minutes.
Downlink data includes:
- An updated frame from the orbital camera.
- Science camera images (one every [ten seconds]) taken since the last
downlink.
- Any telemetry data stored since the last uplink.
Downlink products are automatically populated in the file hierarchy on the
server. See the server documentation for more details on this process.
Commands should be generated and sent to the server in NQC syntax. Although there is no restriction on how frequently this can occur, this will be
limited in practice by the availability of state information contained in the
downlink. If a target is moved by the rover before acquiring a scoring science image, the moved target will no longer be valid (i.e. it will no longer be
possible to score points on that target).
At the end of [fifteen] [two-minute] cycles, game play will end and a subset
of science data products will be submitted for scoring.
Scoring
Judges will assess the set of science data products submitted by the users.
The objective is to have science images in which single target multipliers can
be clearly resolved. Scores for science images are derived by multiplying the
value of the target by the value of the visible multiplier. The value label for a
given target need not be visible in the image, though science data products
should be clearly labeled with which target they came from. Images with
more than one multiplier visible will score zero points. For each target, only
the highest multiplier captured will generate a score (the rest are ignored if
submitted).
|
