Interview with Dr. Rohan Chandra—Winner of Drones 2023 Young Investigator Award

Dr. Rohan Chandra is currently a postdoctoral research fellow at Texas Robotics, advised by Dr. Joydeep Biswas and Dr. Peter Stone, at the University of Texas, Austin, USA. His research focuses on algorithms and systems for enabling robots to navigate safely and efficiently among humans, like humans. Dr. Chandra obtained his M.S. and Ph.D. in 2018 and 2022 from the University of Maryland, USA, advised by Dr. Dinesh Manocha, and completed his B.Tech at Delhi Technological University, New Delhi, India in 2016. His doctoral thesis focused on autonomous driving in dense, heterogeneous, and unstructured traffic environments.

The following is an interview with Dr. Rohan Chandra:

1. Can you tell us a little about your background and what initially sparked your interest in your field of research?

I completed my bachelor’s in India in 2016 in the field of electronics and communication. Following that, I went on to complete my graduate degrees in computer science from the University of Maryland and at UMD. My doctoral dissertation was focused on autonomous driving in challenging highly unstructured environments. I just recently completed a two-year postdoc at the University of Texas at Austin, and now I will be joining the University of Virginia as an assistant professor this fall in the Department of Computer Science. However, my initial interest was not always in robotics. I was always interested in math and problem solving. When I was in grad school, my advisor gave me a thesis topic to work on in the field of autonomous driving. Along the way, I just figured out how to combine my research interest with my passion for mathematical problem solving. This journey led me to make significant progress in the field of autonomous driving and robotics. 

2. Did you face any challenges and, if so, what challenges did you face during your research and how did you overcome them?

One of the challenges in my field, especially in robotics, is the availability of data, because you have to train machine learning models and neural networks based on data. And when I was working on my thesis, which as I mentioned, was on autonomous driving in very challenging traffic environments, some of the most challenging traffic environments were those found in Asian countries, like India, China, or Singapore. Most of the datasets that are available or that were available at the time were on easier traffic environments like those found in the Western parts of the world such as the US or Europe. So, this lack of data presented a huge problem because then you're not able to move forward in your research. To overcome this issue, I traveled to these regions to collect data myself. For example, I flew to Singapore to gather traffic data and collaborated with a company in India to collect data on Indian traffic. Just last year, we published the dataset that we collected in India at a conference. Another challenge in robotics is dealing with real physical systems. Unlike software, real robots can break down, making it difficult to keep them operational. During my graduate studies, I had no experience with physical robots, which is why I pursued a postdoctoral fellowship at the University of Texas. This experience allowed me to gain practical knowledge and hands-on experience with real robots, furthering my expertise in the field.

3. How do you believe open access contributes to the advancement of knowledge and accessibility within your field?

I firmly believe in open access research. There was a recent discussion on social media among prominent scientists and engineers about what defines science and what drives scientific progress. A key point that emerged in that discussion was that science thrives when ideas can collide and challenge each other, which can only happen if those ideas are open and accessible. If you're doing the research behind closed doors, then those ideas are not allowed to be vetted in peer-reviewed venues and it's difficult for science to progress. It is essential for ideas to be vetted, questioned, and challenged repeatedly, which is a part of the definition of open science. In robotics, open research is particularly crucial. Access to code and hardware specifications from previous experiments is vital. Without open infrastructure, researchers risk constantly reinventing the wheel instead of building on prior work. Open access allows for rapid prototyping and progress by enabling researchers to build on existing experiments, which is especially important in the fast-evolving field of robotics.

4. When it comes to your goals, what are your short-term and long-term research goals?

I will answer that question in reverse order, starting with my long-term goals. My long-term goal is to enable mobile robots to collaborate with and assist humans in homes, public spaces (e.g., airports, hospitals, etc.), and transportation. There are several objectives to complete towards achieving this goal. First, we need multiple robots to navigate complex human environments in a fully decentralized manner both safely and efficiently. Second, we need to build a new paradigm for human–robot interaction that can facilitate complex natural language communication between humans and robots. Third, we need to extend autonomous driving towards highly dense, heterogeneous, and unstructured traffic. I have several active projects that fulfill these objectives. 

5. What research topics do you think will be of particular interest to the research community in the coming years?

I’ve been thinking about that, and I believe that it's a great time to be in robotics right now, especially with recent advancements. But also, if you look at deep learning, in deep learning we had the ideas and algorithms all the way from the 1980–1990s. But it was only 12 years ago, in 2012, that the engineering and the infrastructure caught up with the GPUs and then the revolution came. In robotics right now, it’s the other way around, where we are making significant progress in terms of infrastructure, with more GPUs, more data, and more foundational models. However, we are still relying on outdated methodologies, ideologies, and algorithms. What’s needed in robotics now is an algorithmic shift in how we approach manipulation, locomotion, navigation, and human–robot interaction. In the next 5–10 years, alongside the industry's efforts to scale infrastructure, researchers need to develop new algorithmic paradigms in human–robot interaction, multi-robot systems, and robot learning to drive a revolution similar to what deep learning achieved.

6. Can you share any advice for young researchers who are just starting out in your field?

One crucial aspect, especially in robotics, is maintaining strong foundational knowledge and fundamentals. While having AI skills is essential in fields like computer vision or natural language processing, where the focus is on images, text, or speech, robotics requires dealing with real physical systems. Operating these systems effectively demands a solid engineering and mathematical background in addition to AI expertise. To be a leader in robotics, it is important not to solely focus on AI skills but to also ensure a strong foundation in engineering principles. This approach will be invaluable in advancing your career and contributing to the field's progress. 

Please join us in congratulating Dr. Rohan Chandra on winning the Drones 2023 Young Investigator Award!