Knowledge-based Robot Sequential Decision-Making under Uncertainty

Dr. Shiqi Zhang

[Abstract]

Robots need sequential decision-making (SDM) capabilities to complete tasks that require more than one action. Robot SDM is difficult for reasons such as imperfect perception, unreliable action outcomes, and incomplete domain knowledge. This talk will cover a few ways of incorporating declarative knowledge representation and reasoning (KRR) into robot decision-making paradigms such as probabilistic planning and reinforcement learning. The contextual knowledge typically comes from humans, and tends to be sparse but potentially effective. The algorithms are mainly demonstrated using tasks of mobile service robots, such as humanguidance, object delivery, and dialog systems.

[Biography]

Dr. Shiqi Zhang is an assistant professor at the Department of Computer Science, SUNY Binghamton. He received his Ph.D. (2013) in Computer Science from Texas Tech University. Before that, he received his MS (2008) and BS (2006) from Harbin Institute of Technology in China. From 2014 to 2016, he was a Postdoctoral Fellow at the University of Texas at Austin. He co-offered a tutorial on robot decision-making at AAAI-19, and co-chaired the AAAI-18 Spring Symposium on robot goal-directed autonomy. He received a Best Robotics Paper Award from AAMAS-18. His research lies in the intersection of artificial intelligence and robotics. He is particularly interested in developing reasoning and sequential decision-making algorithms to help robots achieve long-term autonomy in human-inhabited, collaborative environments.