| Summary | The assignment guides students through the creation of a planning domain. The primary language is PDDL (syntax commonly used for planning), and visualization+testing elements are included. |
| Topics | Automated Planning |
| Audience | Undergraduate AI course |
| Difficulty | Medium. Preliminary understanding of PDDL and planning are required. |
| Strengths | The problem setting is engaging, and allows students to explore various aspects of planning encodings. |
| Weaknesses | Assignment works best when instructors can help diagnose the services used. |
| Dependencies | Basic Automated Planning understanding, and knowledge of PDDL syntax. Web browser is all that's needed to work on the assignment. |
| Variants | Variations on the mechanics of the domain, or the marking schema, have been explored across different instructors. |
There are two versions of this assignment, run at two academic institutions on opposite ends of the globe. Both are provided to give a sense of how the assignment is delivered.
Version 1
Version 2
- Description (zip of markdown and images)
- Template Files: domain problem1 problem2 problem3
Abstract
This assignment is dedicated to testing a student's proficiency in implementing a medium-difficulty domain in the Planning Domain Definition Language (PDDL). It guides the students through a series of problems of increasing difficulty and contains ample scaffolding for them to implement the needed actions. The assignment has been piloted across several undergraduate Artificial Intelligence courses and is the perfect companion to a module that teaches students about Automated Planning. Two versions are provided with this submission. The first uses a traditional grading scheme, while the second encourages ungrading for the students. There are also slight differences in the expected mechanics of the domain and the problem descriptions (multiple variations are helpful so that solutions cannot be readily found online).
The assignment and its variations have been given to thousands of students over the last few years, and it is consistently ranked as one of the best (if not the top) among the assignments used in undergraduate AI classes. We are excited to offer this to other AI instructors, hoping it can be used more broadly.
Sample solutions are available for instructors upon request, as well as other modules to learn about other PDDL features and more expressive planning modeling languages.