IEEE Robotics & Automation Magazine - December 2017 - 128

Figure 2. The finalists and co-organizers of HRATC 2017 in Singapore at ICRA.

mapping (SLAM) techniques. However, these solutions came mostly in
the form of standard ROS packages
designed for indoor operation and,
thus, performed poorly in the highly
dynamic outdoor scenario. As there
was no reliable information on the
robot position, the teams had difficulties navigating the arena and were
unable to detect a satisfactory number
of mines. The organizing team realized
that the testing area did not have a sufficient number of unique features for
the teams' SLAM algorithms and so
augmented it with artificial obstacles. However, the problem of obstacle
detection and avoidance became
aggravated, as some teams could not
detect the artificial obstacles and
would crash into them, ending their
test run. This issue was mainly caused
by all teams using only the top laser
rangefinder and not using the downward-tilting rangefinder placed specifically to detect smaller obstacles. Teams
primarily did this because most navigation software, such as the ROS navigation stack, does not require this
second laser, and the teams decided
not to use it because it would require
major changes in the software.
After the simulation and testing
phases of the challenge, three teams
qualified as finalists: Team Dhruva
(India), Team National University of
Singapore, and Team RCMakers (Turkey). At the finals, held in conjunction
with the International Conference on
128

IEEE ROBOTICS & AUTOMATION MAGAZINE

Robotics and Automation (ICRA) in
Singapore, teams were asked to send in
their code to be run on the remotely
stationed robot in Brazil (see Figure 2
for a photo of the finalists and the coorganizers of HRATC 2017). After two
trials for each team, it was decided that
there would be no winner this year
because, according to the rules of the
challenge, a winning team is required to
reliably detect more than 50% of the
number of buried mines.
After four years of the HRATC, we
undertook an honest self-assessment to
evaluate how well the challenge is fulfilling its original objectives. During these
four years, a total of 53 teams representing 18 countries from the Americas,
Europe, and Asia have participated. The
most positive aspects have been 1)
increasing the awareness of the robotics
community for the humanitarian demining problem and the possibility of
addressing it using robots in a costeffective fashion as a viable alternative
and 2) engaging students at the educational level toward developing a robust
framework supporting cooperative
remote field robotics trials.
The HRATC is the only mine detection event in the world that provides
free access to a state-of-the-art robot
platform and sensors to participants
from all over the globe, thereby promoting the education of roboticists from
developing countries with advanced
tools. This has resulted in significantly
lowering the entry barrier for students

DECEMBER 2017

and researchers, who otherwise do not
have a practicable way to participate in
an international robotics challenge
geared toward the benefit of humanity.
Thanks to its sponsors, IEEE-RAS
SIGHT and the IEEE ICRA Challenges
Committee, travel support has been
provided for the challenge's finalists so
they could attend a premier robotics
conference and network with attendees
of the ICRA.
In terms of measuring the success
of the event from a technical point of
view, the inclusion of additional sensors for determining the robot's
ground-truth would be beneficial as it
could be used to evaluate the quality of
the teams' localization solutions. There
is a tradeoff between realism and
accessibility in the testing phase, which
the organizing team had to consider
when selecting a test arena. An intermediate localization phase could be
useful for the teams to focus on and
solve localization issues before the testing phase. In this phase, the organizers
could provide data sets of all the robot
sensors for several trajectories, including its ground-truth. The teams would
then implement and enhance their
localization solutions, and those having the positioning error below a certain threshold would be allowed to
proceed to the testing phase. In the
current setup, the laser rangefinder
proved to be insufficient for the selected scenario as the robot could not
detect enough features for localization
purposes. Perhaps a more affordable
sensor, such as Sweep Scanner [6], could
be a better option due to its reduced
cost and longer range.
We are considering the development of performance indices that
would be independent of the environment so the progress of the teams
could be compared year after year,
even if the operating environment is
drastically altered. Continual adjustments will be made to ensure more
sustainable progress in the challenge's
outcomes and eventually reach the
point where state-of-the-art advances
can be quantified, leading to practical
deployment. One of the most challenging aspects for the teams has

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2017