AI Planning for Consistency Repair
| Vortragende(r) | Jing Wang | |
|---|---|---|
| Vortragstyp | Bachelorarbeit | |
| Betreuer(in) | Bowen Jiang | |
| Termin | Fr 21. August 2026, 14:00 (Raum 010 (Gebäude 50.34)) | |
| Vortragssprache | Englisch | |
| Vortragsmodus | in Präsenz | |
| Kurzfassung | Modern software systems are often described by multiple interconnected models. Changes to one model may cause inconsistencies if related models are not updated. Existing consistency management approaches can propagate predefined changes, but are often limited to fixed repair routines and cannot optimize between repair alternatives or handle global constraints.
This thesis presents a generic planning framework for automated model consistency repair. It extracts a semantic representation from input models using an Abstract Syntax Tree (AST), combines it with consistency rules and global constraints, and translates the repair problem into the Planning Domain Definition Language (PDDL). AI planners then generate repair plans that restore consistency while considering optimization objectives such as repair cost. By separating model semantics from planning representations, the framework can be reused across different modeling languages. A prototype demonstrates the complete workflow and its integration with the Vitruvius model consistency framework. | |