About Us
We are proud to be part of the Computer Science and Engineering (CSE), University at Buffalo (UB).
At the intersection of perception, spatial reasoning, and decision-making, our long-term research goal is to endow mobile robots with human-level autonomy. This vision drives us to develop algorithms and systems that enable robots to efficiently and robustly:
- Perceive and interpret various sensory inputs such as images, point clouds, and proprioceptive data.
- Integrate neural and symbolic representations of spatial common sense and semantic knowledge.
- Reason and plan in real time to navigate, interact, and adapt within unstructured and dynamic environments.
Our current work focuses on establishing the scientific foundations to empower robots to perform the above three tasks with robustness, adaptability, reliability, and transparency by integrating data-efficient learning with structured reasoning. A central paradigm we advocate is imperative learning (IL) - a unified neuro-symbolic learning framework for robot autonomy. It is designed to be self-supervised, modular, and end-to-end trainable, seamlessly combining the flexibility of data-driven models with the generalizability of structured reasoning, e.g., symbolic reasoning with physical laws and logical rules.