Summary of Safe Learning Of Pddl Domains with Conditional Effects — Extended Version, by Argaman Mordoch et al.
Safe Learning of PDDL Domains with Conditional Effects – Extended Version
by Argaman Mordoch, Enrico Scala, Roni Stern, Brendan Juba
First submitted to arxiv on: 22 Mar 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A powerful domain-independent planner is developed for solving various planning problems, requiring a model of the acting agent’s actions. Manually designing such an action model is challenging. An alternative is to automatically learn safe action models from observation, ensuring consistency with real actions. Algorithms exist for learning safe action models, but they struggle with domains featuring conditional or universal effects. This paper proves that learning non-trivial safe action models with conditional effects requires an exponential number of samples. However, under reasonable assumptions, a tractable algorithm called Conditional-SAM is proposed. Theoretical analysis and experimental evaluation show that the learned action models can solve most test set problems in experimented domains. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A planner helps robots make decisions. Creating a model of how the robot will act is hard. Instead, we can learn this model from watching the robot. This is safe if it only creates plans that match what the real robot does. Some planners are good at learning these models, but they struggle when there are special effects in the planning problem. This paper shows that learning these models with special effects needs a huge amount of data. But under some reasonable rules, an algorithm called Conditional-SAM can learn these models. The results show that these learned models can solve many problems. |
Keywords
» Artificial intelligence » Sam
