Model-based Workflow for the Automated Generation of PDDL Descriptions

Hamied Nabizada,Tom Jeleniewski,Felix Gehlhoff,Alexander Fay

2024-08-15

Abstract:Manually creating Planning Domain Definition Language (PDDL) descriptions is difficult, error-prone, and requires extensive expert knowledge. However, this knowledge is already embedded in engineering models and can be reused. Therefore, this contribution presents a comprehensive workflow for the automated generation of PDDL descriptions from integrated system and product models. The proposed workflow leverages Model-Based Systems Engineering (MBSE) to organize and manage system and product information, translating it automatically into PDDL syntax for planning purposes. By connecting system and product models with planning aspects, it ensures that changes in these models are quickly reflected in updated PDDL descriptions, facilitating efficient and adaptable planning processes. The workflow is validated within a use case from aircraft assembly.

Artificial Intelligence,Software Engineering

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that manually creating Planning Domain Definition Language (PDDL) descriptions is difficult, error - prone and requires a great deal of expertise. Although this knowledge has been embedded in engineering models, there is currently a lack of an effective method to reuse this knowledge to automatically generate PDDL descriptions. Therefore, the paper proposes a model - based workflow for automatically generating PDDL descriptions from integrated system and product models, thereby improving the efficiency and adaptability of the planning process.

Specifically, the paper focuses on the following aspects:

1. **Challenges in Manually Creating PDDL Descriptions**: Manually creating PDDL descriptions is not only time - consuming but also error - prone and requires a large amount of expertise.
2. **Reuse of Existing Knowledge**: Engineering models already contain a large amount of information about systems and products, and this information can be reused to generate PDDL descriptions.
3. **Requirement for Automatically Generating PDDL Descriptions**: By automatically generating PDDL descriptions, the time and errors of manual operations can be reduced, and the efficiency of the planning process can be improved.

To solve these problems, the paper proposes a workflow consisting of four stages:

1. **Analysis and Preparation of System Models**: Analyze existing system models, define the scope of observation, and identify elements related to the current planning problem.
2. **Enrichment of System Models**: Embed planning - specific annotations and actions in the system model to generate an extended system model.
3. **Provision of Product Models**: Extract relevant information from the product model and annotate it into a format conforming to the PDDL domain to generate an extended product model.
4. **Generation of PDDL Descriptions**: Automatically generate PDDL description files, including PDDL domain and problem files, from the extended system and product models.

Through this workflow, the paper aims to achieve an automatic conversion from system and product models to PDDL descriptions, thereby simplifying the planning process of complex systems.