BU ENG Researcher Wenchao Li Receives Prestigious NSF CAREER Award
By Maureen Stanton
Wenchao Li, an ENG assistant professor of electrical and computer engineering with affiliate appointments in systems engineering and computer science, has won a prestigious Faculty Early Career Development (CAREER) Award from the National Science Foundation (NSF).
Li’s research sits at the intersection of formal methods and machine learning, with a focus on building safe and trustworthy autonomous systems. This new five-year grant will support Li’s research on a new machine-learning paradigm called specification-guided imitation learning (IL).
IL is a learning approach where machines, such as robots or artificial intelligence (AI) systems, learn to mimic human actions from demonstrations provided by human experts or expert agents. While IL has proven invaluable for a host of applications – from gaming to self-driving cars, robotic surgery and more – learning from human demonstrations poses a number of challenges.
“Human demonstrations can be inadequate, partial, imperfect, environment-specific, or suboptimal in practice,” explains Dr. Li, director of the Dependable Computing Laboratory and faculty affiliate of the Hariri Institute for Computing, and BU Center for Information & Systems Engineering.
Instead of completely relying on demonstrations, Li’s new NSF project will develop a novel framework where the data-driven learning process of IL is guided by theories and algorithms that employ formal specifications.
“Our approach aims to overcome limitations of learning by demonstration approaches, such as over-reliance on data and lack of environment adaptivity,” says Li.
Li says that the central idea behind this project is that learning from both specifications (i.e. what the task entails) and expert demonstrations (i.e. how the task can be done) can be far more effective than learning from demonstrations alone.
“It is more natural and economical for an expert to specify safety properties that the learned policy must satisfy, instead of showing the many different ways that the system can fail,” explains Li. “Imagine crashing an expensive vehicle hundreds of times just to generate such ‘negative examples’.”
Li’s research will explore what types of specifications would be appropriate for different application scenarios and how to incorporate them into a data-driven learning process like imitation learning.
Li’s NSF CAREER project will also help lay the foundation for the next generation of scientists by providing outreach opportunities to K-12 students from disadvantaged backgrounds, broadening the participation of women and minorities in STEM research, and helping inspire younger generations to pursue a college degree or a future career in engineering.
“This NSF CAREER grant explores the synergies between data and formal specifications in a unified learning framework, and has the potential to transform how we develop AI systems in the future,” says Li.