Refining Process Descriptions from Execution Data in Hybrid Planning Domain Models

Abstract

The creation and maintenance of a domain model is a well recognised bottleneck in the use of automated planning; indeed, ensuring a planning engine is fed with an accurate model of an application is essential in order that generated plans are effective. Engineering domain models using a hybrid representation is particularly challenging as it requires accurately describing continuous processes, which can have complex numeric effects. In this work we consider the problem of the refinement of an engineered hybrid domain model, to more accurately capture the effect of the underlying processes. Our approach exploits the information content of the original model, utilising machine learning techniques to identify important situation and temporal features that indicate a variation in the original effect. We use the problem of modelling traffic flows in an Urban Traffic Management setting as a case study and demonstrate in our evaluation that the refined domain models provide more accurate simulation, which can lead to higher quality plans. The contribution of this work is a general approach to the automated refinement of hybrid planning domain models that reduces the knowledge engineering effort in producing a detailed process model. The approach can be used for refining the domain model during the initial stages of development, or for re-configuring the domain model when used in the same problem area but with a different scenario. We test out the approach within a real world case study.

Towards Automated Modeling Assistance: An Efficient Approach for Repairing Flawed Planning Domains

“…Additionally, Sreedharan et al (2020) studied how to deal with an incorrect dialogue domain via techniques developed for explainable AI planning Kambhampati 2020), e.g., via model reconciliation (Sreedharan, Chakraborti, andKambhampati 2021;Sreedharan, Bercher, and Kambhampati 2022). Lindsay et al (2020) proposed an approach based upon machine learning which refines an inaccurate hybrid domain model. Göbelbecker et al (2010) and Gragera, García-Olaya, and Fernández (2022) investigated approaches which turn an unsolvable planning problem into a solvable one.…”

Improving Task Planning Knowledge Robustness for Autonomous Robots

“…The RACE system (Rockel et al, 2013) employs HTN planner to learn from experiences, while ProbCog (Karapinar et al, 2012) employs logic programming (ILP) and geometric reasoning module. The work in (Lindsay et al, 2020) refined hybrid domain models. It utilises ML techniques to identify the serious situation and temporal features.…”

“…The work in (Lindsay et al, 2020) considered the problem of refining hybrid domain models. It proposed a general approach to the automated refinement of such domain models.…”

Was Fixing This Really That Hard? On the Complexity of Correcting HTN Domains

“…To our best knowledge only a few approaches exist that provide further AI-based support. Lindsay et al (2020), for example, refined an inaccurate hybrid domain to capture the environment more accurately, and Sreedharan et al (2020) revised a dialogue domain via model reconciliation (Sreedharan, Chakraborti, and Kambhampati 2021; Sreedharan, Bercher, and Kambhampati 2022). Lin and Bercher (2021) also studied the complexity of finding corrections to a flawed domain model provided a plan that shall be a solution but currently is not.…”