SPIRE : systems for plan recovery in intelligent environments

Abstract

As intelligent spaces have become more prevalent, the need for diagnosis and recovery from failures in these spaces has also grown. In an effort to make these types of spaces more useable for people who neither care about nor want to understand the underlying technology, encapsulation of high-level ideas into plans is being explored. Unfortunately, these plans are not always executed flawlessly. To deal with these failures, we have developed a system for plan recovery in intelligent spaces called a desktop multi-player virtual simulation game. In Charles River City, students work in teams to SPRIE. SPRIE is an automated system that uses a two-stage process for recovery from plan execution failures. In the first stage, the failure is diagnosed using Bayesian networks. In the second stage, a alternate plan is selected to recover from the failure. By automating the recovery process, the burden of understand the underlying architecture of the intelligent space is removed from the user.