Neuro-Symbolic AI for Unified Robot Autonomy

Robot autonomy spans perception, planning, and control, yet these components are often developed in isolation. Is there a principled paradigm that unifies them within a single learning and reasoning framework? In this talk, Dr. Wang addresses this question by introducing Imperative Learning, a self-supervised neuro-symbolic framework for robot autonomy. Imperative Learning is formulated as a specialized bilevel optimization framework that integrates three tightly coupled components: a neural module, a symbolic reasoning engine, and a structured memory system. This design enables reciprocal learning across modules, combining the representational power of neural networks with the structure, consistency, and interpretability of symbolic reasoning. Dr. Wang will show how this unified framework supports a broad spectrum of autonomy tasks, including localization and mapping, control and motion planning, high-level task planning, vision-language navigation, etc. Extensive empirical results across diverse robotic platforms show that Imperative Learning significantly strengthens robustness, generalization, and data efficiency, offering a coherent foundation for the next generation of autonomous systems. 

Chen Wang is an Assistant Professor in the Department of Computer Science and Engineering at the University at Buffalo (UB), where he leads the Spatial AI and Robotics (SAIR) Lab. He received the BEng degree from the Beijing Institute of Technology (BIT), China, and the PhD degree from Nanyang Technological University (NTU), Singapore. His research focuses on perception, spatial reasoning, and decision-making for mobile robots, with the goal of advancing robot autonomy toward human-level capabilities.