University of California, Berkeley
Liting Sun is a Ph.D. candidate in the Department of Mechanical Engineering, University of California at Berkeley. Prior to that, she received a B.S. degree in mechanical engineering from the University of Science and Technology of China in 2009 and visited UC Berkeley as a visiting student researcher and a junior specialist from 2012-2015. She was the recipient of the Chinese National Encouragement Scholarship in 2007 and 2008. She was also awarded the J. K. Zee fellowship and D. GALE fellowship in Berkeley in 2016 and 2018. Her research interests lie in intelligent and high-performance behavior design for interactive autonomous systems, merging ideas from robotics, optimization, control, behavior economics, game theory, machine learning, and artificial intelligence. Applications of her researches include autonomous vehicles, robotics, and high-precision motion systems. Besides technical aspects, Liting has been very active in mentoring and academic activities. She was awarded the Outstanding Graduate Student Instructor in Berkeley in 2018. She also co-organized the workshop “Prediction and Decision Making for Socially Interactive Autonomous Driving” in 2019 IEEE Intelligent Vehicles Symposium, and the workshop “Benchmark and Dataset for Probabilistic Prediction of Interactive Human Behavior” in 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems.
Intelligent and high-performance autonomous systems such as collaborative robots and autonomous vehicles have been well recognized for their potential of improving the safety and efficiency in many application domains, from manufacturing to transportation, and from work to daily life. To ensure that their interaction with human is safe, efficient and reliable in a dynamic environment, their behavior design has to satisfy requirements from three aspects: 1) intelligent behavior design so that they can understand the complex human behavior which is in nature hierarchical, casual but irrational, versatile and full of uncertainties; 2) socially compliant behavior design to allow them smoothly interact with human; and 3) reliable and high-performance behavior design to guarantee the performance under uncertainties. To tackle these challenges, my work first explored interpretable behavior and cognitive models to describe human behavior. Inverse reinforcement learning approaches with the theory of mind, cumulative prospect theory and game theory were developed to help the autonomous systems effectively understand what the human intends to do and how they act. Based on that, learning from demonstration and optimization-integrated imitation learning were designed to enable more human-like behavior in terms of social norms and skills such as courtesy and social perception. Finally, for reliable and high-performance execution behavior, advanced control with adaptation and iterative learning rules was developed, which eliminated the influence of environmental uncertainties by considering their different characteristics. Applications on autonomous vehicles, robot manipulators and high-precision motion systems were also addressed.