IEEE Robotics & Automation Magazine - September 2018 - 101

ROI-8 and increased the importance of ROI-7, which was
turned down before the competition due to the existence of
large boulders in the region. We also shifted the center of ROI-7
away from the boulder location to reduce the collision risk.
This effort, however, failed, as Cataglyphis searched too far
away from the actual ROI-7 location. At 100 min into the challenge, a second communication update was made to move the
ROI-7 location closer to the boulders. After that, Cataglyphis
found and returned the easy sample in ROI-8 and found another hard sample in ROI-7. However, upon collecting the hard
sample, the camera cast a shadow over the sample and caused
Cataglyphis to abandon it shortly before the end of the 2 h run.
Overall, during the final NASA SRR Challenge, Cataglyphis performed 2,692 m of autonomous driving, searched
17 ROIs (including 11 unique ROIs), and returned five samples. A playback of the SRR challenge using recorded data can
be seen in our video [9].
Lessons Learned
The main takeaway from developing and testing Cataglyphis was
that the physical world has many different ways to cause a robot
to fail, and we, the human designers, are not yet capable of predicting them all (see Figure 9). Fundamentally solving this issue
requires new ways of thinking to enable more flexible robot
autonomy architecture that can deal with surprises. We also
learned that interacting with the physical environment was very
beneficial in guiding our thinking process. By directly working
with a robot and by repeatedly testing it in new, relevant environments, new ideas and inspirations were constantly generated.
Conclusions
Robotic foraging is a rich and potentially fruitful research field.
The successful performance of Cataglyphis during the NASA
SRR challenge demonstrated that robust robotic foraging in a
benign outdoor environment is now a reality. With additional
development, several practical applications, such as pollination,
pest control, and street trash removal, may soon be performed
by robots. The general framework that enabled autonomous
sample return as described in this article may also be adapted to
robots performing these applications. To facilitate this process,
the source code for Cataglyphis is shared on our repository [10].
Acknowledgments
The research presented in this article was funded by West Virginia University, the West Virginia Space Grant Consortium,
and the West Virginia High Technology Consortium Foundation. We thank Analog Devices, Netburner, AutonomouStuff,
Advanced Circuits, MidWest Motion Products, GitHub,
Robotshop, and NVIDIA for their support.
References
[1] E. Nilsen, C. Whetsel, R. Mattingly, and L. May, "Mars sample return
campaign status," in Proc. 2012 IEEE Aerospace Conf., pp. 1-7.
[2] Y. Gu, N. Ohi, K. Lassak, J. Strader, L. Kogan, A. Hypes, S. Harper, B. Hu,
M. Gramlich, R. Kavi, R. Watson, and J. Gross, "Cataglyphis: An autonomous sample return rover," J. Field Robot., vol. 35, no. 2, pp. 248-274, 2018.

[3] A. Das, M. Diu, N. Mathew, C. Scharfenberger, J. Servos, A. Wong, J.
S. Zelek, D. A. Clausi, and S. L. Waslander, "Mapping, planning, and
sample detection strategies for autonomous exploration," J. Field
Robot., vol. 31, no. 1, pp. 75-106, 2014.
[4] S. Rusinkiewicz and M. Levoy, "Efficient variants of the ICP algorithm," in Proc. 3rd Int. Conf. 3-D Digital Imaging and Modeling, 2001,
pp. 145-152.
[5] R. Kümmerle, G. Grisetti, H. Strasdat, K. Konolige, and W. Burgard,
"g2o: A general framework for graph optimization," in Proc. 2011 IEEE
Int. Conf. Robotics and Automation (ICRA), pp. 3607-3613.
[6] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "Imagenet classification with deep convolutional neural networks," in Proc. 25th Int. Conf.
Neural Information Processing Systems, 2012, pp. 1097-1105.
[7] A. Valero-Gomez, J. V. Gomez, S. Garrido, and L. Moreno, "The
path to efficiency: Fast marching method for safer, more efficient
mobile robot trajectories," IEEE Robot. Autom. Mag., vol. 20, no. 4,
pp. 111-120, 2013.
[8] T. M. Howard, M. Pivtoraiko, R. A. Knepper, and A. Kelly, "Modelpredictive motion planning: Several key developments for autonomous
mobile robots," IEEE Robot. Autom. Mag., vol. 21, no. 1, pp. 64-73, 2014.
[9] West Virginia University Interactive Robotics Laboratory. (2016, Sept.
13). Cataglyphis rover at final NASA Sample Return Robot Centennial
Challenge. [Online]. Available: https://www.youtube.com/watch?
v=ThKvmDHuXdU
[10] West Virginia University Interactive Robotics Laboratory. (2016). Cataglyphis onboard software. [Online]. Available: https://github.com/
wvu-irl/cataglyphis3-software

Yu Gu, West Virginia University, Morgantown, United States.
E-mail: yu.gu@mail.wvu.edu.
Jared Strader, West Virginia University, Morgantown, United
States. E-mail: jared.l.strader@gmail.com.
Nicholas Ohi, West Virginia University, Morgantown, United
States. E-mail: nohi@mix.wvu.edu.
Scott Harper, West Virginia University, Morgantown, United
States. E-mail: sharper9@mix.wvu.edu.
Kyle Lassak, West Virginia University, Morgantown, United
States. E-mail: klassak@mix.wvu.edu.
Chizhao Yang, West Virginia University, Morgantown, United
States. E-mail: cy0003@mix.wvu.edu.
Lisa Kogan, West Virginia University, Morgantown, United
States. E-mail: likogan@mix.wvu.edu.
Boyi Hu, West Virginia University, Morgantown, United
States. E-mail: bohu@mix.wvu.edu.
Matthew Gramlich, West Virginia University, Morgantown,
United States. E-mail: mpgramlich@mix.wvu.edu.
Rahul Kavi, West Virginia University, Morgantown, United
States. E-mail: rkavi@mix.wvu.edu.
Jason Gross, West Virginia University, Morgantown, United
States. E-mail: jason.gross@mail.wvu.edu.

september 2018

IEEE ROBOTICS & AUTOMATION MAGAZINE

101

https://www.youtube.com/watch?v=ThKvmDHuXdU https://www.github.com/wvu-irl/cataglyphis3-software

IEEE Robotics & Automation Magazine - September 2018

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