Neuro-Symbolic Skill Discovery for Conditional Multi-Level Planning

Researchers have developed a new AI architecture that learns high-level symbolic skills from minimal low-level demonstrations, enabling robots to manipulate objects and execute complex tasks in unseen environments. The system combines neural networks for symbol discovery with visual language models for high-level planning and gradient-based methods for low-level execution.

• →The architecture can learn generalizable skills from just a few unlabeled trajectory demonstrations without extensive training data.
• →Visual language models are integrated to automatically interpret discovered action symbols and generate high-level plans.
• →The system successfully manipulates objects in previously unseen locations and cluttered environments.
• →The method preserves both high-level symbolic reasoning and low-level action planning capabilities.
• →Real-world experiments demonstrate the ability to execute long-horizon tasks using novel action sequences.