Procedure Planning

Procedure planning focuses on automatically generating sequences of actions to achieve a desired outcome, often based on visual input like instructional videos or medical imaging. Current research emphasizes developing robust models, such as diffusion models and transformer-based architectures, that can handle uncertainty, learn from limited supervision (e.g., weak or text-based labels), and incorporate external knowledge sources like commonsense reasoning or procedural knowledge graphs. This field is significant for advancing AI capabilities in areas like robotics, medical automation, and human-computer interaction, enabling more efficient and reliable automation of complex tasks.