Efficient context free parsing of multi-agent activities for team and plan recognition

Authors:
Bikramjit Banerjee;Jeremy Lyle;Landon Kraemer
Affiliations:
The University of Southern, Mississippi, Hattiesburg, MS;The University of Southern, Mississippi, Hattiesburg, MS;The University of Southern, Mississippi, Hattiesburg, MS
Venue:
Proceedings of the 11th International Conference on Autonomous Agents and Multiagent Systems - Volume 3
Year:
2012

Citing 3
Cited 0

HTN planning: complexity and expressivity

AAAI'94 Proceedings of the twelfth national conference on Artificial intelligence (vol. 2)
Getting serious about parsing plans: a grammatical analysis of plan recognition

AAAI'90 Proceedings of the eighth National conference on Artificial intelligence - Volume 1
Multi-agent plan recognition with partial team traces and plan libraries

IJCAI'11 Proceedings of the Twenty-Second international joint conference on Artificial Intelligence - Volume Volume One

Quantified Score

Hi-index	0.00

Visualization

Abstract

We extend a recent formalization of multi-agent plan recognition (MAPR), to accommodate compact multi-agent plan libraries in the form of context free grammars (CFG), incomplete plan executions, and uncertainty in the observation trace. Some existing approaches for single agent plan recognition cast it as a problem of parsing a single agent activity trace. With the help of our multi-agent CFG, we do the same for MAPR. However, known hardness results from multi-agent plan recognition constrain our options for efficient parsing, but we claim that static teams are a necessary (though not sufficient) condition for polynomial parsing. The necessity is supported by the fact that MAPR becomes NP-complete when teams can change dynamically. For sufficiency, we impose additional restrictions and claim that if the social structure among the agents is of certain types, then polynomial time parsing is possible.