AI Planning for Consistency Repair
| Typ | Bachelorarbeit oder Masterarbeit | |
|---|---|---|
| Aushang | AIPlanningForConsistencyRepair.pdf | |
| Betreuer | Wenden Sie sich bei Interesse oder Fragen bitte an: Benedikt Jutz (E-Mail: benedikt.jutz@kit.edu, Telefon: +49-721-608-45995), Bowen Jiang (E-Mail: bowen.jiang@kit.edu) |
Motivation
The Vitruvius framework manages multiple models in a Virtual Single Underlying Model (V-SUM). Model Consistency is required to build software from these models. Therefore, Vitruvius maintains constructive, change-driven consistency by applying Consistency Preservation Rules (CPRs) in response to changes made by modellers.
Because maintaining consistency at all times is infeasible, Vitruvius needs to tolerate inconsistencies, and resolve them later. Tolerating inconsistencies becomes possible with a declarative notion of consistency, where consistency relations are expressed in a predicate-like manner. However, resolving such inconsistencies is not possible right now.
Task Description
In this thesis, you will apply Automated Planning, a sub-discipline of symbolic AI, to restore declarative consistency for V-SUMs. The following steps are required to do so:
- Express inconsistent V-SUMs as a planning problem in a suitable language, such as the Planning Domain Definition Language (PDDL).
- Transform V-SUMs into planning problems, and solution plans to changes in the Vitruvius framework.
- Evaluate whether CPRs are applicable to restore declarative consistency.
- Evaluate the scalability of solvers for automated planning with regards to time and memory constraints.
Note that the workload can be reduced, should the topic prove to be too extensive.