A Visual Integrated Development Environment for Automated Planning Domain Models

Abstract

Automated planning software uses symbolic reasoning techniques and models of the planning domain to generate plans. Execution systems execute these plans to perform tasks or achieve goals, subject to constraints imposed by physical laws, resource limits, and environmental conditions and flight rules. Automated planning can support autonomous operations of spacecraft, habitats, and space launch systems. The Action Notation Modeling Language (ANML) is a relatively new language developed by NASA for specifying planning domain models. Developing and maintaining good planning domain models is challenging and critical to the success of applying applying automated planning technology to support autonomous systems. We developed an integrated development environment (IDE) to help modelers enter, review, test, debug, maintain, and enhance ANML planning domain models as well as review and understand ANML models developed by others. The IDE is implemented in the Java programming language using the Eclipse and Xtext open source frameworks for developing integrated development environments (IDEs). The IDE provides a syntax-aware text-based editor that color-codes ANML text based on its syntactic type, flags errors and warnings, supports browsing, suggests code completions, and provides online help. It automatically detects and highlights problems such as syntax errors, type mismatches, references to undefined variables, and incorrect numbers or types of arguments in references to variables and actions. In addition, the IDE provides graphical displays that help modelers see important patterns and relationships among planning variables and actions. An evaluation showed that PM/IDE significantly reduced the time needed to create ANML models. The syntax highlighting and tooltips helped modelers avoid many syntax errors, and the visualizations helped modelers identify logic errors and other semantic issues.