Foundations of Human-Agent Collaboration:
Situation-Relevant Information Sharing
There is a pressing need for the development of new techniques in collaborative systems for representing and reasoning about joint task achievement in dynamic environments. As automated systems become more sophisticated in their capabilities, the design of effective interaction with human operators becomes more demanding (e.g., the recent Toyota Announcement).
and reasoning about joint task achievement.
This project tackles a challenging problem faced in collaborative systems. Human-agent robotic teamwork, also termed human-automation teamwork, involving teams comprised of software agents, robots and humans, is increasingly being exploited to carry out tasks such as remote management of air or ground vehicles, and robot-assisted search and rescue operations. Such use of software assistants and physical robots to support human activities will increase in coming years. When action outcomes can be uncertain, successful collaborative activity cannot be fully pre-scripted, but must allow for adjustment as events unfold.
The goal of this project is to discover new algorithms and software prototypes that will support the development of human-automation teams that can coordinate and collaborate in fast changing task environments.
Multi-agent Epistemic Planning With Proper Doxastic Knowledge Bases
Research on epistemic planning implemented in the Hattari board game
Video demo of the online Hattari game
Video of our multi-agent planner, MA-PRP, running on a Nao playing Tic-Tac-Toe against a person
Related Venues and Groups
PhD applicants invited to apply for 2014/5 - the agentlab presently has a range of exciting research opportunities for graduates to do their thesis in the area of intelligent agents - funded by APA, APAI, Melbourne University scholarships and NICTA top-up scholarships and DSI (STRAPA) scholarships, please contact Adrian Pearce for more information if you are interested in applying.
|Funding source||ARC Grant DP130102825|
|Project time frame||2013 - 2015|