The Toronto Intelligent Decision Engineering Laboratory
Execution Monitoring

Execution Monitoring of Rich Plan Representations


Members

J. Christopher Beck
Sheila McIlraith
Christian Muise

Project description

The area of automated planning focuses primarily on the generation of plans for a given domain and problem. However, when a plan is generated for execution by an agent, one must take care that the continued execution remain valid and ultimately achieves the goal. This is the focus of Execution Monitoring.

Our work focuses on developing methods for monitoring the execution of rich planning representations. By monitoring the execution of a more expressive plan representation than a classical sequence of actions, our approach encompasses many more possibilities for achieving the goal. Succinctly representing a policy for the continued execution of the plan is at the core of our approach, and by strategically building the policy we effectively unify the task of state evaluation (determining if our plan continues to be valid) and action selection (picking the best action to get us closer to our goal).

Start date

February 1st, 2009

Funding

NSERC CGS-D 2009-2012

Publications

  1. Monitoring the Execution of Partial-Order Plans via Regression. Muise, C.; McIlraith, S. A.; and Beck, J. C. 2011. In International Joint Conference On Artificial Intelligence

University of Toronto Mechanical and Information Engineering