An Agentic Framework for Proactive Robots with Procedural Assistance Memory
This work proposes a memory-based framework for proactive robots to anticipate user needs by learning recurring routines. It uses a Vision-Language Model (VLM) to predict tasks based on stored Procedural Assistance Memory (PAM), which is updated daily through self-reflection. This system consistently improves performance, reducing response delays significantly and adapting to changes in user habits without needing retraining.