Peter Jarvis, Ian Miguel and Qiang Shen
We argued in our Flexible Graphplan (FGP) work that the classical definition of the planning problem is too rigid to capture the full subtlety of many real problems. In light of this, we provided a new flexible definition and described a solution strategy based upon the Graphplan framework. Under this definition an action must determine how well its preconditions are met and assert a relative satisfaction degree along with its effects. In this paper, we describe how the Blackbox framework can be modified to solve the same flexible problem. Our Flexible Blackbox (FBB) system can synthesise a range of plans for a given flexible problem, trading the compromises made in a plan versus plan length in the same manner as FGP. We detail the modifications required and provide empirical results that compare our implementation of FBB with our FGP solver on a range of flexible problems and against vanilla Blackbox, STAN, IPP, and Graphplan on imperative problems.