Research Archives - Karla Stepanova

Research mobility at the University of Freiburg

This month I had the opportunity to join the 𝗥𝗼𝗯𝗼𝘁𝗶𝗰 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗚𝗿𝗼𝘂𝗽 at the University of Freiburg, led by Abhinav Valada (continuing the robotics tradition established by Wolfram Burgard). Building 80 became my home for one month of research and exchange.

It is difficult to express what makes going abroad so essential for research, but it simply is. It cannot be replaced by online meetings, lectures, or even conferences. Being immersed in a new environment changes how you think, opens new perspectives, and gives you capabilities you simply can’t unlock otherwise. For me, it feels like moving to the next level of a game—new rules, new features, new challanges and new possibilities.

I got to study many new things, discuss the same problems from fresh perspectives, and work intensively on one problem for a full month.

And what did I already gain from this stay?
• New insights and approaches to research problems
• The opportunity to focus deeply on one topic for a month
• Discussions with outstanding researchers (most of them part of the ELLIS PhD program)
• A view into how another group organizes its communication, labs, robots, meetings, computational resources, retreats, knowledge exchange, etc.

It wasn’t easy to organize mobility with a family, but it worked out: the university provided us with a reasonably priced flat, and my child even can spend a month in a German forest kindergarten 🌲. Plus, Freiburg is a really cool small city in Germany—perfect for outdoor activities. It is close to the French and Swiss borders, cyclist-friendly, and surrounded by hills. And there is a direct night train between Prague and Freiburg :).

💡 Think research mobility is a waste of time? Just try it—you’ll see how much you gain from it.

Grateful to #ROBOPROX, the University of Freiburg, prof. Valada and to Adrian Roefer, for making this experience possible.

Below are some of my more detailed insights from the mobility.

Highlights from Freiburg – Week 1

I’m really happy that I had the chance to spend eight hours a day back at the computer programming again. Most of my work here is with Adrian Röfer, who did his master’s with Michael Beetz.

On the very day I arrived, there was the PhD defense of Daniel Hollerkamp, the first finished PhD student of Prof. Abhinav Valada as a head of the RL group and of the Freiburg Robot Learning Group itself (news link). After 5.5 years, Daniel’s achievements were not bad: 6 RAL papers, 1 TRO paper, 1 patent, and multiple conference publications including some best paper awards (Google Scholar).

Robotics in Freiburg has long been associated with Wolfram Burgard (link) who is now in Norimberg. Amusingly, I almost got to use his very old bike during my stay.

The group also has a weekly team lunch on Thursdays, which was this week moved to the PhD defense celebration… I am happy that Michal is keeping pushing me to have the Thursday lunch together…let’s keep the tradition.

The Robot Learning Lab has an open repository policy—all of their repositories are openly shared within their GitLab to group members.

In terms of topics, the lab focuses mainly on two topics: autonomous driving (Prof. Valada’s original specialization) and robot manipulation. Like many other labs, they also have a demo apartment setup—a two-room flat with a kitchen and living area stocked with fruit and vegetables—where Frankas on Clearpath platforms move around, VR setup, and garage with mobile robots and a car full of sensors. They also have a PR2 robot (though only one arm works) and a KUKA iiwa, which unfortunately has currently some issues.

The University of Freiburg also provides apartments for visiting researchers, which is how I managed to find a reasonably priced place to stay for a month. I think this would be a fantastic idea to adopt at CTU too.

As for the city itself—Freiburg is a true paradise for cyclists. Everyone cycles here, and bikes are given as much space and respect as cars (or maybe more). There are even dedicated bike roads, and on shared ones, half the road is usually reserved for cyclists. And this is considered “quiet season” since the semester, schools, and kindergartens haven’t started yet. Even so, it already felt like way more cyclists than in Prague. I wish Prague could move more in this direction.

Highlights from Freiburg – week 2

Freiburg is a nice flat city surrounded by hills and forests. Perfect for cycling on city bike in the city and hiking in the hills around :). Ok and there is a nice place called university, where interesting research happens as well ;-).

Even in the place where there are very bright minds and great discussion and cooperation, people always feel that something could be better. As one of the PhD students said “The grass is always greener on the other side”. And that is why it is good to go abroad. To feel amazed by the things they do well and after a while to also see the places where they do not do that well.

However, I have to appreciate the drive of the students here. They were working on ICRA submissions last week (basically everyone worked towards a paper, only some didn’t submit). All of them discuss and know what they want to show and what does it mean to have ready to publish results. They know how to compare and which metrics to evaluate. And if they don’t achieve the good enough results, they don’t submit and submit a month later to RAL, you can feel the confidence they feel that they can do it, it is only when. I didn’t see any decline in their focus on the paper after some of them decided not to publish in ICRA and plan to do it only few week later to RAL with ready results.

What surprised me, is that they have basically no homeoffices. Maybe they can ask once a month for a day off, but otherwise, they are expected to be at work every day.

On Wednesday last week my phone started to beep and get read “National emergency alarm”. The test of sirens which we have in Czech republic every month here happens only once or twice a year and it is not only sirens, but all the mobile phones in the Germany. And they have a good reason for testing it as 3 years ago or so I heard there was an emergency but the whole system didn’t work at all…

Everyday is time to learn new things. I got into understanding stochastic clique matching for matching various objects across demos, and got to revisit and thinking my old symbol grounding problem. I spent a nice afternoon and evening thinking about mapping language and vision by probabilistic bipartite matching and probabilistic assignment to improve over methods that I tried during my PhD. Could I actually improve compared to CLIP by utilizing these graph methods? Well, many people obviously tried several of these things in annotating visual images accompanied by weak language annotations. But still I feel there is a space in the area. Let’s see if I get back to it, or it just stays on the backyard of ideas that have never been realized…

I like the rounded monitor I got borrowed :).

Highlights from Freiburg – week 3

After two weeks of digging through the code line by line, exploring its structure, and experimenting in iPython notebooks, it felt great to finally make my first commit — even if it was just a data structure for handling input/output. This week, I made more and more of my own commits and started contributing by adding new functionality. It feels good to actually produce some code, though I still feel insecure. I know I’m not a great software engineer and don’t understand many things, but that’s exactly why I’m here — to learn and be evaluated. Every day I get the chance to feel like an idiot, and I like it. If it’s not too much, it moves you forward.

I had to laugh when Adrian said he’s having a “publication crisis” — despite already having four first-author papers and around 13–15 total as he nears the end of his PhD. When I said that doesn’t sound bad, he replied: “Well, we always evaluate ourselves based on those we have around us. And one thing Abhinav is really good at is selecting people. I always feel I get to work with excellent people.” I couldn’t agree more. A researcher should always feel a bit “not good enough” — it means you’re surrounded by those who challenge you and push you forward.

After the weekend, the temperature dropped by almost 20 degrees, with rainy autumn days replacing the heat. My 45-minute walks to university are now wet, shoes squelching “quatch quatch,” but I couldn’t survive 10+ hours at a desk without them.

The managerial responsibilities from Prague, although reduced (thx everyone for that!), didn’t vanish: organizing ICDL, project proposals and meetings, teaching prep, and students. But with Jan’s support here, I learned to manage it by getting up at 5 AM, walking to the office, and spending the early hours on Prague tasks before everyone arrives (also nice for privacy, since I share an office with four others — a real challenge for me). From 9 AM I focus on the project I came here to do. The cost: by 2 PM, my brain is often drained. The first weeks’ motivation carried me, but afternoons are tougher now. At least going to bed early with Robert gives me enough sleep.

I’m still not used to sitting and thinking so long in a row. Strange thoughts creep in: that I’ve become a “mind machine,” my body irrelevant in this setup. But our minds evolved to support our bodies, embodied by nature. Will we really reduce ourselves to minds (so easily replaced by artificial ones), or keep our bodies central? My walks and weekends feel like resistance against that pull. Over lunch, when we compared Czech and German focus on beer, someone joked: “Alcohol is so popular here because it’s the only fun left — Germans just work, go home, save money… but it’s changing slowly.” It actually fit nicely with my recent listening to Lex Fridman’s podcast with Norman Ohler on drug use in the German army during WWII (from pervitin in pharmacies to meth enabling blitzkrieg marches, and much more).

The lab itself has a nice culture. Most communication goes through Slack, group meetings happen every two weeks. They also have weekly lunches together, time to time retreats, and occasional pub outings (mostly organized by PhDs and postdocs themselves). You can see as the lab is new (5 years), they try to find their way. There’s even a new “fun coordinator” and mini-golf in the hall. PhD students teach for four semesters and take ownership of projects early on. Professor steps back after proposals, having weekly discussions with each student. He focuses more on new students (on their first papers and projects) and later gives PhDs more and more independence and less support with the usefulness of weekly meetings depending on their topic.

By the end of last week, the building was nearly empty — most people had left for CoRL in Korea, some continuing directly to IROS in China or taking vacations. By the time they return, I’ll already be gone.

I can only recommend going abroad. It brings more than you expect — especially if you leave as many things as possible behind.

Highlights from Freiburg – week 4

This was my last week in Freiburg. My stay is slowly coming to the end. Makes me feel a bit sad. On Friday, I went out for a beer with Adrian — we talked about life and research, about what it means to become a professor in Germany and what that really involves. I even shed a few tears thinking about how wonderful my time in Freiburg has been, and how much I wish research could always feel like this: focusing deeply on a problem, moving things forward, and being able to dedicate enough time to real research and programming. Somehow, in Freiburg, it worked. Now the challenge is how to learn to have the same in Prague.

It was also our last weekend in Freiburg. We took a cable car up to Schauinsland and went for a beautiful walk through the hills surrounding the city. Mist and drizzle, autumn knocking on the door. I really like this place — a city flat enough to bike through effortlessly, yet with hills rising a thousand meters right behind your house in every direction. Not bad at all. We also stopped at the giant swing overlooking Freiburg. In a small nearby village, at a bus stop, I saw a sign that said “Hitchhiking Stop.” As someone who grew up hitchhiking and still enjoys it, it made me so happy to see that it’s coming back. The bus runs only once an hour there, so maybe the locals are encouraging each other to give someone a ride.

At work, it’s quiet. Adrian left for vacation, and I managed to work on our joint research for another two days before giving in and switching to writing my ERC proposal. I’d been postponing it as long as I could, but in an almost empty lab, it was hard to stay motivated.

One day, Nicolas went out to buy wood — why? They’re building a wall where they’ll hang all the successfully defended PhD theses. They told me it is a tradition coming from Sweden. Now that the first defense is done, it’s the right time for them to start the wall. Seems like a very nice idea.

In the city center, we came across wooden boxes full of bread and a couple of old bikes with crates of vegetables beside them. We found out that there are about ten of these “food-saving” stations around Freiburg. Individuals, bakeries, and shops leave food they can’t use anymore, and anyone passing by can take what they need. Jan talked to the man who maintains the boxes — he said the system has been running for a while.

And there’s another nice local habit: on weekends, people put things they no longer need outside their homes — clothes, furniture, dishes, books — a simple, neighborly exchange of unneeded things. Whatever’s left unclaimed just gets taken to the recycling center. We picked up some children’s games and even found a swimming goggles for our kid.

On the last day of September, we started clearing out the apartment and giving away the last of our things. I packed up the rest, spent the night, had one last breakfast, loaded up my bike bags, and rode to the university in the morning. On the way, I dropped off leftover food — oil, unopened pasta, and few more things — at one of the food-saving boxes and kept going. I bought Zwetschgenkuchen as a farewell cake for the few people who weren’t away at CoRL, and in the evening, I rode my bike toward the train station.

When I washed my mug and glass, closed the lab door, got on my bike, and left the university behind, I felt a wave of nostalgia. In just a month, this place had already grown close to my heart. Not as deeply as when I spent five months in Plymouth, but still — I truly liked it here. And I can’t wait to travel again to another place.

Presenting Research Isn’t Easy—Here’s What I Did Wrong

Yesterday, I presented our paper in a seminar, which went terribly. It wasted my colleagues’ time and failed to communicate the essence of our work. I felt frustrated about the wasted opportunity to present our work well and sorry for those who attended. I wanted to reflect on what went wrong and how I could improve.

So, what went wrong with my presentation? I see at least three main issues. First, even though I knew the paper well, I didn’t have enough time to refresh all the details before the seminar. Second, due to my partial involvement, I was not an expert on every aspect of the paper—only certain parts—yet I felt obligated to present all sections equally. Finally, and most importantly, I failed to think through what I wanted to communicate to the audience.

As a result, I started the seminar and quickly became frustrated with my own presentation. I felt unprepared, which led to a mental fog, increased anxiety, and a spiral of self-doubt. Under stress, I struggled to articulate the motivation behind our work, let alone explain the paper’s contributions clearly. Ultimately, I couldn’t even present the material I had prepared.

This post is my reflection on what went wrong and, more importantly, how I can improve for future presentations.

Lack of Thorough Preparation

Regarding the first point – thorough preparation. I know I struggle with unprepared speech. In general, I express myself much better in writing or when I am thoroughly prepared, which helps me manage the stress of social anxiety. That’s why I need enough time to organize my thoughts and ensure I am well-prepared before speaking publicly. When I teach, I every year again recompute exercises beforehand to ensure I can answer questions confidently. I make the concepts so clear in my mind that even under stress, I can discuss them effectively. This takes a lot of time.

Similarly, preparing for a seminar requires a significant investment of time. Although I had allocated enough time, I failed to use it effectively due to ad hoc requests—family demands, “unexpected” tasks, and difficulty saying no to last-minute requests from professors and students. None of these are excuses—I should have accounted for these variables. I failed to schedule my preparation well in advance, so the things to which I cannot say no (like my son demanding me in kindergarten for carnival) are not having such a big effect on preparation.

Presenting a collaborative work that I didn’t entirely own

The second issue was presenting work I didn’t do entirely myself. Even with thorough preparation, I won’t know every detail of a collaborative project. There will always be unexpected questions I can’t answer, and that’s okay. However, confidence and preparation are crucial to handling this without panic. The key is to present the parts I do understand effectively. The more multidisciplinary and collaborative research I engage in, the less I can be an expert in every aspect of the work. Retaining details of things I haven’t personally done requires additional effort.

My long-term memory isn’t great—I even forget parts of my own work after some time, let alone the work of others. With increasing responsibilities across multiple disciplines and roles—psychology, neuroscience, AI, robotics, managing people, purchases, teaching, leading student works on various topics, preparing projects, communicating with companies, accepting new PhD students, attracting girls to robotics, preparing study programs,…—the demands on my memory exceed my capacity. Again, refreshing my knowledge well before every single event and paper presentation is simply necessary. I know I need more preparation time than some others to make everything click. Public speaking confidence improves with practice, even for me, but thorough preparation is essential for me to manage it and not get overwhelmed by the stress.

Many professors deliver excellent talks on various topics without doing the whole work—often even better than those who fully understand every detail. How is that possible? First, they are likely smarter and have way better memory than I do. Second, they focus on delivering a clear message, emphasizing key points they understand well, and diving into details where they feel confident.

Failure to define a clear message

This ties closely to the last issue – the importance of crafting a clear message. Instead of just reviewing the paper and describing the experiments, I should have asked: What makes this seminar worth attending? What should attendees learn? A deep dive into a specific algorithm? Gain insights into method comparisons? A better grasp of experimental design? What are the best practices for collecting a strong evaluation dataset?

A good seminar leaves the audience with at least one meaningful insight. I should have identified a single key takeaway and explored it deeply instead of covering everything superficially. I must underline the main message and go in-depth on the parts I understand well. I have to accept that I will never understand everything, but I should still be able to explain at least something really well—and unfortunately, that was not the case in this seminar. Next time, I need to carefully evaluate what I know and don’t know in advance and focus deeply on the aspects where I can provide real value.

Learning to say no

I’m not a science rock star who can give an amazing talk with minimal preparation. I know this, and that knowledge increased my stress before the presentation even began—especially when I knew I hadn’t prepared as I would have wished to and I saw an unexpectedly large audience.

I also felt a strong sense of responsibility—especially in my new role as a group leader, where I set the standard for quality. If I announce a seminar, I need to deliver a talk that is at least worth attending. While perfection isn’t necessary, the seminar must provide value—just as teaching must be worthwhile for students. But how do you make good decisions in an endless flow of tasks and responsibilities where time estimation is crucial but really difficult? One key lesson: I must learn to say NO more often. If I say YES to something, I need to ensure I dedicate the necessary time and effort to do it well. Otherwise, I’ll feel overly stressed and fail to deliver value to others.

Final thoughts

Despite the poor presentation, I received valuable feedback on dataset improvements and feature selection. I’m grateful for this—thank you to everyone who contributed.

To those who attended and questioned why they came—I can’t promise my next seminar will be perfect, but I will strive to do better.

3 months as a group leader – Researcher’s diary #1

It has been three months since I became the head of the Robotics Perception Group (ROP) at CIIRC CTU. In this post, I want to share the challenges I’ve encountered after starting to lead the group and the initiatives I’ve introduced.

Our group

The group consists of 11 researchers, 5 PhD students, 2 technicians/programmers, 1 project manager and 1 admin. We are engaged in various industrial and research projects, focusing on merging classical perception techniques (especially vision) and knowledge representation with deep learning. Our interests include integrating data from multiple modalities to develop robust perception and task understanding for human-robot interaction, as well as industrial applications like welding and automotive perception.

Taking on the leadership of this group has been a significant challenge, and I aim to harness its potential carefully. The team is exceptionally diverse, with members ranging from psychologists and neurobiologists to physicists, mathematicians, and engineers. They span the spectrum from rigorous theoreticians to application-focused researchers. This diversity, cultivated by Prof. Hlaváč, the group’s founder and former head, offers immense potential for interdisciplinary collaboration. However, it also presents communication challenges. Bridging the different “languages” and approaches within the team is not always straightforward, but I appreciate the richness this diversity brings and effort everyone is willing to put to the group.

In these first three months, I’ve identified several challenges, most of them revolving around finding the right balance.

Finding a right balance – the biggest challenge I’ve encountered

Infrastructure vs. Research Freedom
Balancing the need to build infrastructure (e.g., website, GitLab, Wiki, seminars) to facilitate collaboration and sharing with providing enough space for research is tricky. While infrastructure is vital for leveraging our group’s diversity, maintaining it consumes time that could be spent on research, which is our primary output.
Industrial Applications vs. Publishing
Encouraging researchers focused on industrial applications to publish their work is particularly challenging. These projects often demand full attention, and publications are not required. There doesn’t exist a quick fix solution but it rather needs long time effort. I am exploring ways to motivate and support this process more effectively. To address this, I plan to enforce publication commitments as part of industrial project submissions wherever possible. This should improve a balance between application-focused work and academic dissemination.
Personal Vision vs. Research Diversity
I strive to push forward my vision while preserving the group’s diversity. A cohesive vision is essential, but I believe large collaborative projects aligned with my vision can unify us. To this end, I’ve been dedicating significant effort to writing new proposals, including an EU proposal, GACR Junior STAR, and possibly in the close future an ERC Starting Grant.
Managing vs. Researching
Balancing management responsibilities with my own research remains the most challenging aspect of my role. While I engage weekly with bachelor’s, master’s, and PhD students on research topics, I deeply miss the uninterrupted focus on my own projects. As someone who has a limited capacity for social interaction and struggles to focus amidst the presence of others and the distractions of small tasks, constant engagement can be exhausting. Working from home a few days a week has helped to improve this balance. However, it still doesn’t feel entirely right. I know I need to put in more effort to refine this equilibrium and find a way to truly enjoy both research and leadership as the head of the group.

Initiatives introduced

We have introduced several initiatives to improve the group’s functioning:

A weekly newsletter summarizes past events and upcoming activities, keeping everyone, especially PhD students, informed.
Regular group board meetings to discuss topics such as finances, teaching, social activities, dissemination, publications, and collaborations.
Starting in January, we’ll hold a regular machine learning study group, an idea we’ve long discussed and are now implementing.
I encourage team members to focus on their research careers, improve their CVs, and refine their research topics. I discuss the state of their research, suggest areas for improvement, and share insights from my own approach. Whether this effort succeeds will take time to evaluate.
We are collectively enhancing our infrastructure, such as improving the group’s website, creating a Wiki for sharing information, or starting to use LinkedIn to showcase results.

Final remarks

Despite these challenges, I deeply enjoy the responsibility and diversity this position brings. My hope is that everyone in our group feels valued and empowered to become the best version of themselves as researchers.

10 tips to improve your resilience in research

Last month I listened to Hillary Clinton talking about resilience. That gave me the chance to think about how I find resilience important in research and in life in general (doing research with three kids just requires resilience by default :)). And why I also like to build resilience in my kids.

I think that you really have to be resilient if you want to succeed in research. There will always be challenges that you have to face. There will be experiments that go wrong, rejected papers, or someone publishing your ideas before you. You never know what will be the world in next week, next month, next year. To manage this world, you have to believe in yourself and your abilities, know what is your goal, learn how to ask for help, and find always new directions for research.

So what the resilience actually is? Resilience refers to one’s ability to survive and flourish despite the risk, to adjust and adapt to changes as they occur, and to recover from misfortunes. As such, it covers various personal features such as self-belief (confidence in your own abilities), optimism (the belief that you will experience good outcomes in your life), purposeful direction (to which extent you have clear goals), adaptability (how much you are willing to adapt your behavior/approach in response to changing circumstances), ingenuity (how much you perceive that you are capable of finding solutions to problems), challenge orientation (how much you enjoy experiences that challenge you), emotion regulation (ability to remain calm in stressful situations), or support seeking (how much you are willing to ask for help and support in difficult situations).

Here are 10 tips from me on how you can support resilience during your research career:

🎯 Be organized. It might sound simple, but I have a feeling this is where success really starts. At least for me. And after seeing many many students, also for many others. Having always a list of tasks you want to work on, have your results and data saved and reusable, and have noted down your ideas. This all gives you a lot of self-confidence and calmness. Someone asks you to remake experiments or recompute something? You know you can do it. Does a new task come in? You don’t lose track of other tasks. You know what you can afford to postpone, what you have to prioritize. Find your way to keep track of your life. I like to use Trello, a paper list of the current tasks, and a paper notebook to make notes from meetings or note down my ideas.
🎯 Finish your tasks. Do you also like the feeling that comes after you managed to submit a paper before the deadline, defend successfully your project, or after you finished slides for the lecture before the lecture itself? I enjoy it a lot. It boosts my confidence in stars. There is nothing better to build your resilience than the feeling that you can manage your work and tasks. Managing tasks does not mean being perfect. You have to learn what is “good enough”. Try your best to make it as good as possible, but don’t give up and finalize it.
🎯 Find easy wins. To get the feeling that you can succeed, it really helps to find places, where it is easy for you to win. Start with a local conference, with workshops on good conferences, or join the team (not being the first author) that submits a paper to a great journal/conference to learn what is the required quality of the paper and what their work looks like.
🎯 Challenge yourself. Don’t be satisfied with the easy wins. Small successes will encourage you and are really important, but try always to reach for something better. Who said you cannot publish in Nature? Maybe not yet, but you can aim for it. Always find a bit better journal in your field than last time, try to reach out to the great researchers in the field and start collaborating with them, or try to ask companies for collaboration. Maybe some will not answer, maybe you get rejected, maybe you are not ready yet. But maybe they will answer and give you a chance.
🎯 Try again and again. Knowing “what is good enough” is something you have to learn by trial and error. You have to try to submit your papers to several journals and conferences to find out what is good enough for the given journal. To know what is already worth to be published. Proficiency comes from practice. Writing is just another skill that you have to learn. Trying, again and again, will help you to build resilience. You will improve your communication skills, you will learn what people around want and how you can help them.
🎯 Be prepared. I find out, that it really helps my confidence when I am well prepared. If I am preparing for teaching, presenting my work at a conference, talking to potential collaborating companies, or even to students. Preparing yourself, trying your presentation in advance, studying the materials, and trying to ask yourself the questions they might ask. All that helps me to get confidence and avoid a lot of stress from unexpected questions.
🎯 Learn to work smart and hard. I really like to use my energy and power efficiently (see the crow on the front image). With three kids, it is almost a necessity. You have to learn how to read your papers on playgrounds, how to make meetings while the small baby is sleeping on your lap, and how to work at night. Kids also teach you, that you don’t have infinite time for your tasks, so you have to accept imperfections. Don’t be lazy, work as much as you can, but also learn how to recharge your energy, how to relax, and how to reuse what you already have. For me, it really works a lot to be well slept and have enough sports activity to deal with stressful situations and negative feedback. Not that it would be easy, but it is easier.
🎯 Rejection is just feedback. I think that this cannot be repeated enough times. Reviewers are (most of the time) not bad guys. They try their best to give you valuable feedback. If they rejected your paper, it might mean that you didn’t present well your contributions, that your work is not ready yet, or maybe you didn’t well compare it to other works in the field. Take it, try to learn from it, evaluate what is valuable in the feedback, improve your work and submit it as soon as you can again.
🎯 Give yourself deadlines. There is a sad story about one grant proposal, that I work on already for almost two years. The problem is, that that type of grant can be submitted anytime. It has high potential, and the topic is interesting. We submitted several proposals in between and some were already successful. But this one has no deadline and is still on the table. It is really hard for me to work without deadlines. Maybe you have it as well. The problem is, that when you have undone things, it feels very similar to the unsuccessful ones. If anyhow possible, find grants, conferences, or journals with deadlines. If there are no deadlines, try to set your own deadlines. Maybe give yourself some presents, if you make it to the deadline :).
🎯 Seek support. One of the biggest problems I see with my students is that they don’t ask questions. They don’t want to show me the undone work. But when being in trouble, help is the most valuable. Also by explaining your problem to someone else, you will get way more clarity about the problem. Ask Google and if you don’t get your answer, seek help so that you are not wasting unnecessary time.

Have fun with your research!

If you would like to receive my occasional newsletter with tips for researchers and my insights about the research (1-2x/month), sign up for the newsletter here: Sign up for the Researcher’s log newsletter.

Otherwise, you might follow me on LinkedIn: Follow me on LinkedIn

If you are interested in the next run of the course “Research and publish with ease” (Here is a post about the first run), or you know anyone who might benefit from it, you can sign up for the waiting list here: Sign up for the waiting list.

Robot self-calibration – 3 new papers published

I am happy to announce that this year, we got accepted another 3 papers (1 journal and 2 conference papers) which deal with robot self-calibration.

This long term project is running under great supervision of Mgr. Matej Hoffmann, Ph.D. (see his website) who focuses on how babies develop the sense of their bodies and space around it and how self-touch might contribute to it (GACR project, EXPRO project). This connects closely also to self-calibration of human who might learn about kinematics of their bodies via touching its individual parts.

In our efforts we try to transfer these ideas to the robots and explore if the usage of this new sensorical modality might be useful for their calibration. We focus mainly on humanoid robots, but explore also the usefulness of these methods for the calibration of industrial robots.

Our long journey was this year completed by 3 papers – 2 conference papers at Humanoids conference and 1 journal paper in Robotics and Computer-Integrated Manufacturing (RCIM) journal.

Matej Hoffmann prepared a nice summary of the whole project on his website: Automatic robot self-calibration.

Stepanova, K., Rozlivek, J., Puciow, F., Krsek, P., Pajdla, T. & Hoffmann, M., “Automatic self-contained calibration of an industrial dual-arm robot with cameras using self-contact, planar constraints, and self-observation.” Robotics and Computer-Integrated Manufacturing73, 102250 (2022). [DOI][Elsevier – OPEN ACCESS pdf][Project webpage][video]

Rozlivek, J., Rustler, L., Stepanova, K., and Hoffmann, M. “Multisensorial robot calibration framework and toolbox.” in ‘Humanoid Robots (Humanoids), IEEE-RAS International Conference on’, pp. 459-466, 2021. GitHub, video, preprint-pdf.

Rustler, L., Potocna, B., Polic, M., Stepanova, K., and Hoffmann, M. “Spatial calibration of whole-body artificial skin on a humanoid robot: comparing self-contact, 3D reconstruction, and CAD-based calibration”, in ‘Humanoid Robots (Humanoids), IEEE-RAS International Conference on’, pp. 445-452, 2021. preprint-pdf, youtube-video.

These papers nicely complement our former paper, where we on simulated iCub humanoid robot explored how using multiple kinematics chain created by self-touch might improve its calibration.

Štěpánová K., Matěj Hoffmann, Tomáš Pajdla. “Robot self-calibration using multiple kinematic chains – a simulation study on the iCub humanoid robot.” RA-L and ICRA (2019), online, project webpage, video

New 3 years project MIRRACLE

I am happy to announce that we got granted a new research project with the title “Multimodal representation of robotic actions and tasks applied to Learning by demonstration”, where I am a principal investigator. The project starts on 1.7.2021 and will last for 3 years. Looking forward for a great research and nice collaborations :).

We are searching for motivated PhD/PostDoc candidate to join our team: http://imitrob.ciirc.cvut.cz/sites/jobs/Mirracle_phd_postdoc.html

Our newly created project webpage (more info will come soon):

http://imitrob.ciirc.cvut.cz/mirracle.php

YouTube – Accepted paper ICRA 2021 (Embodied Reasoning for Manipulation)

Our paper “Embodied Reasoning for Discovering Object Properties via Manipulation” was accepted for ICRA 2021.

Abstract: In this paper we present an integrated system that includes a reasoning from visual and natural language inputs, action and motion planning, executing tasks by a robotic arm, manipulating objects and discovering their properties. The vision to action module recognises the scene with objects and their attributes and analyses enquiries formulated in natural language. It performs multi-modal reasoning and generates a sequence of simple actions that can be executed by the embodied agent. The scene model and action sequence are sent to the planning and execution module that generates motion plan with collision avoidance, simulates the actions as well as executes them by the embodied agent. We extensively use simulated data to train various components of the system which make it more robust to changes in the real environment thus generalise better. We focus on the tabletop scene with objects that can be grasped by our embodied agent, which is 7DoF manipulator with a two-finger gripper. We evaluate the agent on 60 representative queries repeated 3 times (e.g., ‘Check what is on the other side of the soda can’) concerning different objects and tasks in the scene. We perform experiments in simulated and real environment and report the success rate for various components of the system. Our system achieves up to 80.6% success rate on challenging scenes and queries. We also analyse and discuss the challenges that such intelligent embodied system faces.

Paper support materials: http://imitrob.ciirc.cvut.cz/EQA.html

Video can be seen: https://youtu.be/0TREScVil5I