You are here

Novel approaches for supporting plan recognition



TECHNOLOGY AREA(S): Information Systems

OBJECTIVE: DTRA is seeking research in the area of plan recognition from unstructured text sources.

DESCRIPTION: The area of plan recognition encompasses a set of tasks that identifies and relates sequentially time-based observed actions of an entity to a specified objective. Plan recognition presents a complex, multi-discipline challenge to elements of the Department of Defense as it incorporates human factors analysis for anticipatory analysis and modeling the goals of adversaries and machine learning in environments lack of information and uncertainty. This topic is of particular relevance for DTRA in the area of counter-proliferation.

Technical and challenges in plan recognition include: detecting and understanding complex speech acts; inferring changes in goals based on changes in plans; plan detecting based on layering multiple inference layers lossy data from intermittent interruptions; and identifying the temporal conditions. Other technical challenges relate to the generalizability of plan recognition across domains (e.g., plan recognition of a chemical event as opposed to a biological event or sub-events) and the temporal details of events in identifying plan elements and sequentially ordering them in the context of composing an adversary’s plan and characterizing technical progress. At a system-level, technical challenges include the generalizability of the plan recognition systems into to other domains. Technical areas of interest include: Improvements over the state of the art in formal representation (logic-based rules) for plan recognition; natural language processing research related to the identification of task-oriented dialogues and sub-dialogues and understanding speech actions as they pertain to goal-directed behavior; hypothesis generation and co-reference of goal-directed behaviors across multiple data sets of disparate provenance; and abnormalities in plan formulation such as deception or changes in plans due to external stimuli. Respondents would propose novel research topics in response to one, but ideally, multiple technical areas of interest.

PHASE I: Investigate and identify plan recognition approaches and demonstrate proof of concept.

PHASE II: Develop and demonstrate a plan recognition prototype, test against identified plans and labeled data and integrate with current modeling capabilities. Characterize performance levels and assess utility to user-centric tasks in the context of evaluations that involve real data and operationally-relevant scenarios.

PHASE III DUAL USE APPLICATIONS: Plan recognition algorithms demonstrated and proven in operational settings would be valuable in a wide range of potential applications including law enforcement, counter-terrorism and counter international human-trafficking.


  • Carberry, S. (2001). Techniques for plan recognition. User Modeling and User-Adapted Interaction, 11(1-2), 31-48.
  • Litman, D. J., & Allen, J. F. (1987). A plan recognition model for subdialogues in conversations. Cognitive science, 11(2), 163-200.
  • Goldman, R. P., Geib, C. W., Kautz, H., & Asfour, T. (2011). 3.27 Coupling Plan Recognition with Plan Repair for Real-Time Opponent Modeling. Plan Recognition, 19.
  • Schmidt, C. “Introduction to Plan Recognition”. Rutgers University. Retrieved from: accessed on 12/5/2014.

KEYWORDS: NLP, plan recognition, modeling, adversarial planning, logic-based rules, formal representation, goals, intention


US Flag An Official Website of the United States Government