Simultaneous Plan Recognition and Monitoring (SPRAM) for Robot Assistants

Michael Karg (1) , Alexandra Kirsch (1, 2, 3)

1 Institute for Advanced Studies
2 Human-Computer Interaction and Artificial Intelligence
3 Intelligent Autonomous Systems

Personal robots that are useful helpers for humans need to know about common tasks and habits of their human partners to be able to react adequately to human behavior. Thus, knowledge about human task performance becomes an inevitable part of a robotic system that is aimed to work together with humans in human centered environments like a household. But high uncertainties within the robots sensors as well as unpredictable behavior of humans and partial occlusions usually make it hard to achieve certainty about human task execution. We propose a probabilistic module that performs Simultaneous Plan Recognition and Monitoring by maintaining a belief state about human task execution and simultaneously monitoring plans that the human is likely to execute. This way the robot is able to infer possible reactions even if it may not exactly know which plan its human partner is executing. In the application example of a household robot observing a typical morning routine of a person, we show that our module is able recognize and monitor a set of activities of daily living in real time and predict places that the human is likely to visit in near future.

Domaines: Intelligence artificielle [cs.AI]; Robotique [cs.RO]

Michael Karg, Alexandra Kirsch. Simultaneous Plan Recognition and Monitoring (SPRAM) for Robot Assistants. Human Robot Collaboration Workshop 2013, in conjunction with RSS 2013, 2013, Berlin, Germany. ⟨hal-01684313⟩