Report Title: Efficient Robotic Task Planning Combining Large Language Models and Formal Reasoning: From "Coarse Solutions" to Provably Executable Plans

Speaker: Associate Professor Jianmin Ji

Affiliation: University of Science and Technology of China

Organizer: School of Computer Science and Information Engineering

Time: 14:00, Saturday, November 1, 2025

Venue: Room 1602, Block A, Science and Education Building, Feicui Lake Campus

Report Abstract:

Large language models exhibit a certain degree of general problem-solving ability, but their reliability needs to be improved; formal reasoning yields reliable conclusions, but it is difficult to handle large-scale problems. This report focuses on the collaborative mode of large language models (LLM) and formal reasoning (Answer Set Programming, ASP) in cognitive robotic task planning. First, LLM generates a readable "skeleton/coarse solution" plan, which is then compiled into ASP constraints, and a professional solver performs consistency checking and optimal solution synthesis, thus combining general common sense with verifiable reasoning. For the cognitive robotic task planning problem in home environments, a two-stage system CLMASP is formed, enabling plans to start from natural language and eventually result in provably and executable action sequences. CLMASP is verified on the VirtualHome platform and solvers such as clingo/DLV2: the ASP reasoning time is reduced from >2h to <5s, and the plan executability is increased from 7% to 90%, significantly improving efficiency and reliability.

Speaker Profile:

Jianmin Ji is an Associate Professor and Ph.D. Supervisor at the School of Computer Science and Technology, University of Science and Technology of China. He received his Bachelor's and Ph.D. degrees from the University of Science and Technology of China, and served as a Postdoctoral Researcher at the Hong Kong University of Science and Technology, and a Visiting Scholar at the University of Alberta and Carnegie Mellon University. Dr. Jianmin Ji's main research directions include cognitive robotics, autonomous driving and deep reinforcement learning. For more than ten years, he has been responsible for the cognitive and decision-making modules of the "KeJia" and "JiaJia" service robots of the University of Science