Plan-, Goal-, and Intent- Recognition

Last updated on Apr 26, 2023

A socially-intelligent agent or robot must be able to infer the intent and recognize the behavior of others, based on observations of their actions. We have been investigating efficient algorithms for plan-, activity-, and behavior- recognition, in single-robot and multi-robot settings. Specific areas of contribution include:

Overhearing, an approach for recognizing multi-agent plans by listening in on the communications between agents:
- G. A. Kaminka, D. V. Pynadath, and M. Tambe. Monitoring teams by overhearing: A multi-agent plan recognition approach. Journal of Artificial Intelligence Research, 17:83–135, 2002.
- G. Gutnik and G. A. Kaminka. Representing conversations for scalable overhearing. Journal of Artificial Intelligence Research, 25:349–387, 2006.
Symbolic behavior recognition (SBR) is a plan recognition approach that trades space for runtime. The SBR algorithm is to out best knowledge the fastest plan recognition algorithm in the world.
- D. Avrahami-Zilberbrand and G. A. Kaminka. Fast and complete symbolic plan recognition. In Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI-05), pages 653–658, 2005.
Utility-based plan recognition (UPR), is an approach to recognizing activities in terms of the impact on the observer. UPR algorithms efficiently recognize low-likelihood suspicious events, ignored in traditional approaches.
- D. Avrahami-Zilberbrand and G. A. Kaminka. Incorporating observer biases in keyhole plan recognition (efficiently!). In Proceedings of the Twenty-Second National Conference on Artificial Intelligence (AAAI-07), 2007.
- D. Avrahami-Zilberbrand and G. A. Kaminka. Keyhole adversarial plan recognition for recognition of suspicious and anomalous behavior. Plan, Activity, and Intent Recognition, pages 87–121. Morgan Kaufmann, 2014.
- B. Kaluža, G. A. Kaminka, and M. Tambe. Detection of suspicious behavior from a sparse set of multiagent interactions. In Proceedings of the Eleventh International Joint Conference on Autonomous Agents and Multi-Agent Systems (AAMAS-12), 2012.
Mirroring is an exciting approach to plan- and goal- recognition where the agent utilizes its own planning knowledge to generate recognition hypotheses. Mirroring allows plan recognition to be carried out directly in continuous environments, something not possible with other approacches.
- Gal A. Kaminka, Mor Vered, and Noa Agmon. Plan Recognition in Continuous Domains. In Proceedings of the AAAI Conference on Artificial Intelligence , 2018.
- Mor Vered and Gal A. Kaminka. Heuristic Online Goal Recognition in Continuous Domains. In Proceedings of the International Joint Conference on Artificial Intelligence, 2017.