ALPS: An Adaptive Learning and Planning System

摘要

The goal of the ALPS project is to design and prototype a next generation,distributed, real-time, Al planning architecture. The central theme of our research is the real-world planning systems must necessarily be adaptive; that is, they must reconcile themselves to limited environments, and real-time response. We address these concerns with a combination of methods from deliberative planning, machine learning, distributed computing, probabilistic reasoning, plan repair, and reactive planning.