Model Acquisition in the Modern Era


This tutorial will cover some of the landmark methods in the area of planning action model acquisition that our community has produced over the years. From OBSERVER in the early 90’s to the modern forms of action-label-only LOCM techniques, we will cover both the concepts behind these approaches and grounded hands-on examples for attendees to try for themselves.


Target Audience

The content will be accessible to all those with a foundational understanding of classical planning. Familiarity with propositional logic will an asset, but not essential. For the hands-on portion, access to a laptop will also be necessary.


The tutorial will focus on a handful of the most impactful model acquisition techniques over the years, selected from those we have successfully reproduced in the MACQ software framework. This will amount to some subset of OBSERVER, SLAF, ARMS, AMDN, and LOCM. The half-day tutorial will be divided into two equal parts, likely interleaved to continuously offer attendees the opportunity to see the concepts demonstrated on real data.


The first core component will be a traditional tutorial on the model acquisition techniques directly. Initial motivation on the need to analyze discrete sequential data will be given, focusing heavily on application areas such as embodied robotics and business process mining. Following this, the suite of techniques, in increasing difficulty, will be described.


After each module covering a new model acquisition technique, we will turn to hands-on exploration of the method in the MACQ framework. A development environment for using MACQ is already realized through a combination of the MACQ software and the planutils Docker image. This will allow participants to quickly start playing around with the methods they are learning about, with nothing more than a laptop that has Docker installed.

Model Extraction


Christian Muise

Christian is an Assistant Professor at Queen’s University in Kingston, Canada. He completed my PhD under the supervision of Professors Sheila McIlraith and J. Christopher Beck in the area of Automated Planning, with the Knowledge Representation and Reasoning Group at the University of Toronto. Following his PhD, Christian was a post-doc for two years with the University of Melbourne’s Agentlab studying techniques for multi-agent planning with a project on human-agent collaboration, and then subsequently a Research Fellow with the MERS group at MIT’s CSAIL. Just prior to joining Queen’s Christian was a Research Staff Member for two years at the MIT-IBM Watson AI Lab.

Tathagata Chakraborti

Tathagata is a Research Staff Member at IBM Research AI in the AI Composition Lab, Cambridge (MA). His research interests include human-AI interaction, especially planning and collaborative decision-making with humans in the loop, with applications in human-agent teaming and decision support.