Sanjiban Choudhury: Building robots that work seamlessly with humans

Sanjiban Choudhury
  • New Faculty Year: 2022

Assistant Professor, Computer Science

Academic focus: Imitation learning, Decision making, Human-robot interaction

Research summary: My research aims to enable robots to work seamlessly alongside human partners to accomplish everyday tasks. For robots to acquire skills to work with people, they must solve the problem of interaction and imitation. Interactive imitation learning provides a scalable way to implicitly program robots through human demonstrations, interventions, or preferences. To this end, my work focuses on imitation learning, decision making, and human-robot interaction. I am passionate about the fundamental theory and practical algorithms that enable robots to continually interact with both humans and their environment. Much of my research has been deployed on real-world robot systems — mobile manipulators, self-driving cars, and full-scale helicopters.

What inspired you to pursue a career in this field? My interest in robotics began in my undergraduate years when I led a small rag-tag team to the robot soccer world cup. We lost handily, of course, but it was exciting and humbling to see how challenging it was to get robots to perform the simplest of tasks. This observation stayed with me and eventually motivated me to go deep into research and understand the fundamentals. My academic journey took me to different places, from making cars drive themselves to assistive first responders that clear debris. An emerging theme from these varied experiences was how seemingly simple tasks that we humans take for granted can be so hard to program into robots. It's only when you look closely do you realize that we as humans still lack a universal grammar to explain how we reason and act in the world. Going forward, I am excited to keep exploring this fundamental question and coming at it from several different angles. I have been fortunate to have met influential teachers and thinkers who have inspired me along the way, and I hope to empower future generations of roboticists at Cornell.

What are you most looking forward to as a Cornell Engineering faculty member? I think now is an exciting time to be working in AI and robotics. These technologies are increasingly making their way into our daily lives. However, there remain several open questions on how we as a society integrate with such technologies, what innovations add real value to human lives, and important questions around safety and ethics. I am excited to start my faculty career on both teaching and research fronts. I will be teaching a course, Learning for Robot Decision Making, this Fall. In addition to the core technical concepts, I hope to discuss tools for thinking critically and building unified frameworks to navigate the flood of ideas in the field today. On the research front, I am starting my research group, People and Robots Teaching and Learning (PoRTaL), where we will look at problems in robot decision-making and interactive imitation learning to equip robots with skills to work alongside human partners. There is still some way to go before we have everyday robots in our homes. We hope to address some of the fundamental research challenges toward this goal. I look forward to the many deep discussions and collaborations with the wonderful students and colleagues at Cornell.

What do you like to do when you're not working? I have always loved reading fiction, particularly magical realism. Stories have a way of transporting one to a different place and time. A book and a good cup of coffee make for a great afternoon! I also enjoy traveling, swimming, and being outdoors. I am particularly enjoying Ithaca summers of late; there is so much nature and hikes to explore. More recently, I have been channeling some of my creative pursuits towards content creation, like making sketch videos to explain AI and robotics concepts. You can check out my youtube channel here! :) 

 

Sanjiban Choudhury faculty page

People and Robots Teaching and Learning (PoRTaL) group website

More Spotlights