Common Sense Reasoning: This project focuses on developing new knowledge structures and reasoning methods, specifically utilizing evidence theory and abductive reasoning. The goal is to enable robots to draw logical conclusions from incomplete or uncertain observations, allowing them to make “common sense” decisions in newly encountered situations.

Open World Multimodal Perception: This research line aims to move robot perception beyond pre-defined datasets. It focuses on handling both sensory shifts (changes in how sensors perceive the environment) and semantic shifts (new or unknown objects), allowing robots to generalize their perception and recognize what is happening in unseen or dynamic environments.

Open World Task Execution: This project integrates model-based and model-free planning to help robots execute complex actions where data or environment models are limited. It advances semi-predictable task-planning by combining symbolic planning with reinforcement learning, enabling robots to perform complex tasks like perimeter surveillance even when they encounter unpredictable events.