IEEE Robotics & Automation Magazine - December 2023 - 100

Teams were given
garage space and two days
to set-up their systems. The
avatar robots were tested
on a single test track,
shown in Figure 3. Five
operator control rooms
were shared by multiple
teams; thus, an additional
requirement was that
the operator station had
to be moved and set up
quickly. Furthermore,
the avatar robots had to
be untethered, i.e., powered
by batteries and
communicating only via the XPRIZEprovided
Wi-Fi.
The finals testing required the avatar
to complete a mission that consisted of 10
tasks, described in Table 1, resembling a
mission on a distant planet. The tasks had
to be completed in the specified order
and included navigation among obstacles,
multimodal communication with
a human recipient, simple and complex
object manipulation, judging the weight
TABLE 1. Finals tasks.
TASK
Task 1
100%
Task 2
100%
Task 3
100%
Task 4
100%
Task 5
96%
Task 6
79%
Task 7
71%
Task 8
63%
Task 9
42%
Task 10
25%
"
AVATAR ROBOTS
HAD ONE OR TWO
MANIPULATOR ARMS
WITH GRIPPERS OF
VARYING DEXTERITY:
FROM SIMPLE TO
HUMAN-LIKE FIVEFINGER
HANDS.
.
„
of objects, using a power
drill, and discriminating
texture or roughness of
stones. Figure 4 shows
these tasks. The maximum
mission time was
25 min and completion
of all tasks was required
to be eligible for
the
grand prize.
Completion of each
task was worth one point.
Operator and recipient
judges scored their experience
of being present
in the remote space and
of perceiving the presence of the
remote operator, respectively, using the
criteria listed in Table 2. Each criterion
was assessed on a scale of never/poor
(0 points), sometimes/fair (0.5 points),
and always/good (1 point). Thus, up to
five experience points could be earned.
Finals testing began with a scored
qualifying day on which the judges
operated all teams' avatars on the test
course. The two finals testing days were
open to the public, with more than 2,100
visitors and many media representatives
in attendance and thousands watching
the live stream of the first XPRIZE public
testing event since the 2004 Ansari
XPRIZE. The finals live-stream recording
can be seen at https://youtu.be/
lOnV1Go6Op0.
Three teams completed all 10 tasks
on the first test day, with NimbRo in
the lead (15 points), followed by Pollen
Robotics (14.5 points), and AVATRINA
(14.5 points). The top 12 teams
advanced to Day 2. Table 3 reports the
final results. The better of the two scores
was retained, with the completion time
as the tie breaker.
Four
teams
improved
their
task
score on Day 2, three teams had the
same score on both days, and five teams
had a better score on Day 1. The average
change in task score was one point,
indicating a consistent evaluation of the
avatar systems despite different judges,
system changes between the days, and
good or bad luck.
The task completion rates reported
DESCRIPTION
CONNECTIVITY: HUMAN-TO-HUMAN CONNECTION
The avatar robot maneuvers to the mission control desk.
The Avatar reports to the mission commander and introduces
themselves.
The avatar receives the mission details and confirms them with
the mission commander.
The avatar activates a switch which opens the station door.
EXPLORATION: THE NEW ERA OF TRAVEL
The avatar exits the mission control room through the door and
travels across the planet to the next task.
The avatar must identify the full power canisters that are among
empty canisters.
The avatar places the correct canister into the designated slot which
triggers the lighting of the next task zone.
SKILLS TRANSFER: EXPERTISE WITH NO BOUNDARIES
The avatar navigates along the planet's surface to arrive at
the next task.
The avatar must use the drill to remove the door.
The avatar must reach through the barrier to identify the rough
textured rock and retrieve it.
The percentages report task completion rates of the 12 best teams.
in Table 1 indicate that the top 12 teams
solved all four tasks in the connectivity
domain on both days. The first task that
multiple teams failed to complete was
Task 6, which involved discriminating
between objects of different weight. The
drop in completion rates for Tasks 8-10
was caused by two factors: lack of time
and task difficulty. Specifically, grasping
and using the power drill to remove
a screw was successful only 67% of the
time, and identifying a rough stone by
feeling its texture and retrieving it had
only 60% success rate.
The experience scores given by the
operator and recipient judges were high,
with an average score of 4.5 points, a
minimum score of 3.5, and four teams
receiving a perfect score of 5. This result
indicates that, overall, the judges were
quite satisfied with their experience. The
recipient judges rarely deducted half a
point, but the operator judges deducted
points significantly more frequently.
Four teams completed all 10 tasks,
with two receiving perfect experience
scores from both judges. The fastest
team, NimbRo, took an average of only
35 s per task, which is not much slower
100
IEEE ROBOTICS & AUTOMATION MAGAZINE DECEMBER 2023
https://youtu.be/lOnV1Go6Op0 https://youtu.be/lOnV1Go6Op0
IEEE Robotics & Automation Magazine - December 2023

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2023
