Action-model acquisition for planning via transfer learning (Q2453728)

AI planning represents an area of research in artificial intelligence with applications in a variety of domains, including robotics, process planning, web-based information gathering, etc. Many techniques involved in AI planning need action models to acquire sufficient training data to ensure the learned action models are appropriate. This paper deals with a novel algorithmic framework, namely TRAMP (TRansfer learning Action Models for Planning), which seeks to learn action models in the target domain by transferring knowledge from source domains, assuming they are already created and ready to be transferred. After the presentation of TRAMP, the authors empirically evaluate it and, finally, review previous related works and present their future work.