IEEE Robotics & Automation Magazine - September 2018 - 95

(a)

(b)

(c)

The robots were required to be autonomous during the
SRR Challenge. In 2015 and 2016, short, optional communication updates were permitted every 20 min during the challenge runs. This allowed team members to adjust robot
parameters in paused states while the competition clock was
running, simulating the periodic opening of the communication window in a planetary exploration mission.
The competition field was a fenced-off city park with
hilly terrain, trees, small buildings, boulders, and pits. Satellite imagery and topographic maps were provided to the
teams. Many Earth-specific conditions existed during the
SRR Challenge, including rain, strong winds (e.g., the effects
of hurricane Hermine in September 2016), shadows, and
people as well as autumn leaves and trash that could be
mistaken for samples. These conditions, combined with the
large field (approximately the size of four soccer fields), the
2-h time limit, and only a single attempt allowed for each
team per year made the level 2 SRR Challenge difficult.
Cataglyphis General Foraging Strategy
In the three years that we participated in the SRR Challenge,
Cataglyphis was constantly updated to reflect our improved
understanding of the foraging problem. More specifically,
new sensors, such as a three-dimensional (3-D) lidar and a
fisheye camera, were integrated, more refined manipulation
and homing approaches were developed (in 2015), and an
increased number of decisions were made onboard instead
of being preprogramed (in 2016). Figure 3 shows the designs
of Cataglyphis and its homing beacons in 2014, 2015, and
2016, respectively.
The general sample return strategy for Cataglyphis was to
collect and return one sample to the starting platform at a
time. Although it may seem inefficient, this strategy was used
for many reasons. First, the robot itself did not have to return
home at the end of the time limit, which reduced the mission
risk. Second, this approach avoided the extra weight and
complexity needed for onboard sample storage and sterile
handling. Finally, each sample drop-off also allowed Cataglyphis to reset its navigation states using the homing beacon, ensuring precision sample drop-off on the starting
platform as well as better localization performance during
the next foraging trip.

Robot System Overview
An illustration of Cataglyphis' main hardware components in
its 2016 configuration is shown in Figure 4. The robot was
built around a rocker-bogie drivetrain system with a total
mass around 65 kg. Cataglyphis was powered by five 146-Wh
batteries, allowing more than 3 h of autonomous operation.
An underactuated mousetrap grabber [2] was designed for
sample collection and drop-off, tolerating up to 10 cm of sample position estimation and robot control errors.
The main onboard navigation sensors included six wheel
encoders, six Analog Devices ADIS16485 inertial measurement units (IMUs), and a Velodyne VLP-16 3-D lidar. In addition, a mast-mounted ground-facing camera (Canon 5DsR)
with a circular fisheye lens was used for the sample search. An
Advantech MIO-9290 F computer with a 2.3-GHz Intel Core
i7 processor made up the main computing resource.
Cataglyphis' main software modules and their relationships are highlighted in Figure 5. More details about robot
perception and autonomy systems design are provided in the
following sections.
Robot Perception
Cataglyphis needed to understand the relationship between
itself and the surrounding environment to make the right
High-Resolution
Fisheye Camera
3-D Lidar
Batteries
Mousetrap
Grabber

Rocker
Bogie

Computers
and IMUs
Figure 4. The main hardware components of Cataglyphis in its
2016 configuration.

september 2018

IEEE ROBOTICS & AUTOMATION MAGAZINE

IEEE Robotics & Automation Magazine - September 2018

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2018