PDDLGym: Gym Environments from PDDL Problems
This work addresses the need for more relational benchmarks in reinforcement learning and AI planning, facilitating collaboration between these communities, though it is incremental as it builds on existing tools like Gym and PDDL.
The authors tackled the challenge of creating diverse reinforcement learning benchmarks by introducing PDDLGym, a framework that automatically generates OpenAI Gym environments from PDDL domains and problems, resulting in 20 built-in environments with empirical variations in planning and model-learning difficulty.
We present PDDLGym, a framework that automatically constructs OpenAI Gym environments from PDDL domains and problems. Observations and actions in PDDLGym are relational, making the framework particularly well-suited for research in relational reinforcement learning and relational sequential decision-making. PDDLGym is also useful as a generic framework for rapidly building numerous, diverse benchmarks from a concise and familiar specification language. We discuss design decisions and implementation details, and also illustrate empirical variations between the 20 built-in environments in terms of planning and model-learning difficulty. We hope that PDDLGym will facilitate bridge-building between the reinforcement learning community (from which Gym emerged) and the AI planning community (which produced PDDL). We look forward to gathering feedback from all those interested and expanding the set of available environments and features accordingly. Code: https://github.com/tomsilver/pddlgym