# Safe Learning of Lifted Action Models

arXiv: https://arxiv.org/abs/2107.04169 · HTML: https://mlsanity.com/paper/2107.04169/

## Key facts

- Cited by 16 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 30 references that are in the map.
- First posted on arXiv 9 Jul 2021.

## Builds on (most cited first, 5 of 5)

- [Lifelong Multi-Agent Path Finding in Large-Scale Warehouses](https://mlsanity.com/paper/2005.07371/) (2020)
- [Safe Exploration in Finite Markov Decision Processes with Gaussian Processes](https://mlsanity.com/paper/1606.04753/) (2016)
- [Understanding the Power and Limitations of Teaching with Imperfect Knowledge](https://mlsanity.com/paper/2003.09712/) (2020)
- [Efficient, Safe, and Probably Approximately Complete Learning of Action Models](https://mlsanity.com/paper/1705.08961/) (2017)
- [Learning Partially Observable Deterministic Action Models](https://mlsanity.com/paper/1401.3437/) (2014)

## Built on by (most cited first, 16 of 16)

- [Discovering User-Interpretable Capabilities of Black-Box Planning Agents](https://mlsanity.com/paper/2107.13668/) (2021)
- [Goal Recognition as Reinforcement Learning](https://mlsanity.com/paper/2202.06356/) (2022)
- [Planning Domain Model Acquisition from State Traces without Action Parameters](https://mlsanity.com/paper/2402.10726/) (2024)
- [A Domain-Independent Agent Architecture for Adaptive Operation in Evolving Open Worlds](https://mlsanity.com/paper/2306.06272/) (2023)
- [Learning Lifted Action Models from Unsupervised Visual Traces](https://mlsanity.com/paper/2604.19043/) (2026)
- [Safe Learning of PDDL Domains with Conditional Effects -- Extended Version](https://mlsanity.com/paper/2403.15251/) (2024)
- [Planning for Learning Object Properties](https://mlsanity.com/paper/2301.06054/) (2023)
- [On Automating Video Game Regression Testing by Planning and Learning](https://mlsanity.com/paper/2402.12393/) (2024)
- [Action Model Learning with Guarantees](https://mlsanity.com/paper/2404.09631/) (2024)
- [On Generating Monolithic and Model Reconciling Explanations in Probabilistic Scenarios](https://mlsanity.com/paper/2405.19229/) (2024)
- [Integrating Reinforcement Learning, Action Model Learning, and Numeric Planning for Tackling Complex Tasks](https://mlsanity.com/paper/2502.13006/) (2025)
- [Synthesis of Procedural Models for Deterministic Transition Systems](https://mlsanity.com/paper/2307.14368/) (2023)
- [Textual Planning with Explicit Latent Transitions](https://mlsanity.com/paper/2602.04557/) (2026)
- [RAMP: Hybrid DRL for Online Learning of Numeric Action Models](https://mlsanity.com/paper/2604.08685/) (2026)
- [Differentiable Learning of Lifted Action Schemas for Classical Planning](https://mlsanity.com/paper/2605.13282/) (2026)
- [Learning Action Models with Conditional and Quantified Effects via Uncertainty-Guided Exploration](https://mlsanity.com/paper/2608.30955/) (2026)

## Abstract (arXiv)

Creating a domain model, even for classical, domain-independent planning, is a notoriously hard knowledge-engineering task. A natural approach to solve this problem is to learn a domain model from observations. However, model learning approaches frequently do not provide safety guarantees: the learned model may assume actions are applicable when they are not, and may incorrectly capture actions' effects. This may result in generating plans that will fail when executed. In some domains such failures are not acceptable, due to the cost of failure or inability to replan online after failure. In such settings, all learning must be done offline, based on some observations collected, e.g., by some other agents or a human. Through this learning, the task is to generate a plan that is guaranteed to be successful. This is called the model-free planning problem. Prior work proposed an algorithm for solving the model-free planning problem in classical planning. However, they were limited to learning grounded domains, and thus they could not scale. We generalize this prior work and propose the first safe model-free planning algorithm for lifted domains. We prove the correctness of our approach, and provide a statistical analysis showing that the number of trajectories needed to solve future problems with high probability is linear in the potential size of the domain model. We also present experiments on twelve IPC domains showing that our approach is able to learn the real action model in all cases with at most two trajectories.