US 12,292,876 B2
Computer architecture for plan recognition
Christopher William Geib, Minneapolis, MN (US); Scott Ehrlich Friedman, Minneapolis, MN (US); Pavan Kantharaju, Medford, MA (US); and Robert Prescott Goldman, Minneapolis, MN (US)
Assigned to Smart Information Flow Technologies, LLC, Minneapolis, MN (US)
Filed by Smart Information Flow Technologies, LLC, Minneapolis, MN (US)
Filed on Apr. 27, 2023, as Appl. No. 18/140,104.
Claims priority of provisional application 63/341,157, filed on May 12, 2022.
Prior Publication US 2023/0367762 A1, Nov. 16, 2023
Int. Cl. G06F 16/23 (2019.01); G06F 16/215 (2019.01); G06F 16/901 (2019.01)
CPC G06F 16/2365 (2019.01) [G06F 16/215 (2019.01); G06F 16/9024 (2019.01)] 22 Claims
OG exemplary drawing
 
1. A method comprising:
receiving, by a computing machine, a plurality of observations, the plurality of observations including a set of actions and a set of states;
generating, by a plan recognizer engine at the computing machine, an observation data structure, wherein the observation data structure represents causal structures and hierarchical relationships between states and actions in the plurality of observations;
extending, by a planner engine at the computing machine and in accordance with the causal structures and the hierarchical relationships, the observation data structure to include predicted actions that have not yet occurred and are not from the plurality of observations while maintaining a format and a structure of the observation data structure;
reducing, in accordance with consistency rules stored in a memory of the computing machine, the extended observation data structure by:
determining whether a predicted action is consistent with the plurality of observations and the consistency rules; and
maintaining or removing the predicted action based on whether the predicted action is consistent with the plurality of observations and the consistency rules; and
providing an output associated with the reduced observation data structure, wherein the consistency rules comprise logical constraints derived from domain-specific knowledge.