Marlin P. Strub
Roboticist and C++ enthusiast

I accepted a position as a Robotics Engineer at Gravis Robotics, an early-stage startup that automates heavy machinery. I'm part of the autonomy team and primarily work on motion- and mission-planning. I have been following the research done by the team at Gravis for multiple years and am absolutely stoked to start contributing to its commercialization.

Tiago wrote a paper detailing the Exobiology Extant Life Surveyor (EELS) system, with a focus on mission-planning under uncertainty.

The Exobiology Extant Life Surveyor (EELS) team wrote a paper that presents control and planning algorithms for EELS. One of the core ideas of the paper is to decouple shape control from motion control for screw-based gaits. This lets us tackle the control and planning problems with relatively simple algorithms despite the many (>30) degrees of freedom EELS has. The paper was nominated for a best-paper award.

A reference implementation of EIT* (arxiv, doi), the final algorithm of my PhD thesis, is now also publicly available in the Open Motion Planning Library (OMPL). Implementing this has been more of a challenge than anticipated, both technically and in terms of making time for it.

I was promoted to Robotics Technologist at JPL. Fortunately my responsibilities stay the same, so I still get to do hands-on robotics work on the Exobiology Extant Life Surveyor (EELS) and the Sample Recovery Helicopters (SRH) ground mobility effort.

I conducted another field trial with the Exobiology Extant Life Surveyor (EELS) team, again on the Athabasca Glacier. This time we stayed on the ice for three weeks and tested two robots and multiple science instruments. It was even better than last time, as we were able to move more independently on the glacier having had more rope-training in advance.

I coauthored a paper with Wil and Jon that presents Task and Motion Informed Trees (TMIT*; arxiv, doi), an integrated task and motion planning algorithm. TMIT* combines AIT* with PLANET to improve performance when high-level tasks, such as sorting potentially stacked objects, require reasoning about the order in which path-planning problems are to be executed.

I coauthored a paper with Valentin and Jon that presents Effort Informed Roadmaps (EIRM*; arxiv), an multi-query extension to EIT*. EIRM* deliberatively reuses search-effort when solving path-planning problems that are similar to problems it has solved before. This improves performance on successive problems compared to popular single-query and multi-query planners, including EIT*, RRT-Connect, and PRM*.

I accepted a position as a postdoctoral researcher at NASA/JPL. I'm part of the autonomy team for a snake-like robot called EELS that is being built to explore alien worlds. My work focuses on coming up with path planning and control approaches for EELS and includes contributing to the corresponding modules in its C++ autonomy stack.

I successfully defended my PhD thesis at the University of Oxford (Oxford Archive). There are a thousand people to thank. Here I just want to thank Jon for letting me stand on his shoulders these past few years and Nick and Wheeler for the time, energy, and thoughts they invested to help me improve my work.

My first journal paper got accepted at IJRR (arxiv, doi). AIT* and EIT* improve planning performance by leveraging multiple sources of information with an asymmetric bidirectional search. A high-level explanation of what these terms mean is available on youtube.

I conducted a 10-day field trial as part of the Exobiology Extant Life Surveyor (EELS) team at JPL. We tested our sensors and SLAM algorithms on the Athabasca Glacier in the Canadian Rockies. It was my first time on a glacier - what an incredible experience. Our safety team, Direct Action Vertical (DAV), even let us rappel down moulins and crevasses!

Obstacle clearance in search space is an important optimization objective in path planning because it can result in safe paths. This technical report presents admissible heuristics that can improve planning performance when optimizing this objective (arxiv).

The Open Motion Planning Library (OMPL) has been immensely useful in my own research and I'm stoked to join its impressive team of maintainers. Very much looking forward to contributing more to OMPL.

I use Emacs for many things (too many, perhaps). Emacs has org-mode, which is, among other things, an excellent tool for note-taking and journaling. I fixed this tiny bug in org-mode that was exposed by a package that (subjectively) improves the appearance of org-mode by hiding some markup symbols.

I assisted Jon in surveying the field of asymptotically optimal sampling-based path planning algorithms for an invited paper in the Annual Review of Controls, Robotics, and Autonomous Systems (arxiv, annualreviews). The survey cites 150 publications and includes thorough problem definition and assumption sections.

I helped Mike with a paper that presents the autonomy system used on Axel during the field trial last summer. The autonomy system includes a stability analysis that predicts the interaction of Axel's tether with the terrain. My main contribution was to describe how we adapted ABIT* to work with Axel's tether-induced non-Markovian states. You can find an open-access version of it on the ESP website.

ABIT* and AIT* were merged into the Open Motion Planning Library (OMPL). This is my first major contribution to an open-source library. I'm hoping ABIT* and AIT* will be useful for someone.

I implemented a feature that allows using the Flexible Collision Library (FCL) for distance queries in OpenRAVE. This constitutes my first minor contribution to an open-source library and I'm very happy to finally being able to contribute back to the community.

My second paper got accepted at ICRA (arxiv, doi). AIT* improves planning performance with an asymmetric bidirectional search in which both searches continuously inform each other with complementary information. A high-level presentation of these concepts is on youtube.

My first paper got accepted at ICRA (arxiv, doi). ABIT* leverages advanced graph-search techniques, such as inflation and truncation, in sampling-based path planning. This is the algorithm we tested last summer. A high-level presentation of these concepts is on youtube.

Jon and I had the chance to join the NASA Jet Propulsion Laboratory (JPL) on a field trial for Axel in the Mojave Desert. We tested new path planning techniques and supported the JPL team in running science and autonomy experiments.

I joined Jon's Estimation, Search, and Planning Group at the Oxford Robotics Institute to work on path planning for complex systems.

The Swiss national television produced a 4-part documentary about our row across the Atlantic. If your IP is from Switzerland, you can watch it here. It's in Swiss German, but there's also a version of it with English subtitles on youtube. Edit: There's now also a 90-minute version of it as well.

ETH Zürich asked me to write a blog post about rowing the Atlantic, which I happily did. You can read it here.

Swiss Mocean took third place in the 2017 Talisker Whisky Atlantic Challenge and became the first Swiss fours to row any ocean. It took us 30 days, 4 hours, and 59 minutes to row from the Canaries to the Caribbean Islands.

I successfully defended my Master's thesis at ETH Zurich. I'm immensely grateful to Margarita, Fabiola, and Lucas for their guidance throughout my time at the Vision for Robotics Lab.