Great Innovative Ideas are a way to showcase the exciting new research and ideas generated by the computing community.
The following Great Innovative Idea is from Brian Logan (University of Nottingham), John Thangarajah (RMIT University), and Neil Yorke-Smith (American University of Beirut). Their paper called Progressing Intention Progression: A Call for a Goal-Plan Tree Contest was the Blue Sky Ideas Conference Track winner at the Sixteenth International Conference on Autonomous Agents and MutliAgent Systems (AAMAS), May 8-12, 2017 in Sao Paulo, Brazil.
A key problem for an agent with multiple, possibly inconsistent, goals is: “what should I do next”? What to do next can be formalized as the intention progression problem (IPP): what means (i.e., plan) to use to achieve a given goal, and which of the currently adopted plans (i.e., intentions) to progress at the current moment.
This problem is central to autonomous systems and has a long history in computer science, but research on the IPP is fragmented and suffers from a lack of common terminology, data formats, and enabling tools. We propose a Goal-Plan Tree Competition to draw together research communities and to incentivize research on this important problem. User-supplied domain control knowledge in the form of hierarchically structured Goal-Plan Trees (GPTs) is at the heart of a number of approaches to reasoning about action. GPTs form a natural abstraction of the ways in which an agent can achieve its goals, and hence of reasoning about intention progression.
First, the competition necessitates a common toolset for GPTs which is lacking at present. Second, the competition builds connections between CS communities, such as the community around autonomous agents in the Belief-Desire-Intention (BDI) tradition and the community around Hierarchical Task Network (HTN) planning. Third, the competition raises the profile of the IPP and reasoning with GPTs, and pushes forward research.
Brian Logan: My research interests are in the area of agent systems, and span the specification, design and implementation of multi-agent systems, including software architectures, logics for reasoning, and software tools for system building. A key theme of this work is “safe AI”, specifically ensuring that autonomous intelligent agents and multi-agent systems function safely and in accordance with their design objectives.
John Thangarajah: My interests are in Agent Oriented Software Development (how do we build and construct Intelligent Systems), Agent Reasoning (how can programs behave in smart ways), Intelligent Conversation Systems (how can systems interact intelligently satisfying a user’s information needs), Agent Testing (providing assurance that the systems work) and Intelligent Games (how to construct game plays and intelligent characters).
Neil Yorke-Smith: My research aims to help people make decisions in complex situations. In applying artificial intelligence and operations research to practical problems, my aim is to assist the human decision maker in socio-technical systems.
Brian Logan: artificial intelligence and (multi-)agent systems
John Thangarajah: artificial intelligence and (multi-)agent systems
Neil Yorke-Smith: agent-based planning and simulation
Brian Logan: http://www.cs.nott.ac.uk/~pszbsl
John Thangarajah: http://john.agentprojects.com
Neil Yorke-Smith: http://www.aub.edu.lb/~nysmith