Recent research in AI planning is increasingly focused on enhancing the efficiency and adaptability of planning systems through innovative modeling techniques. A notable trend is the shift from traditional symbolic representations to learned transition models, which allow planners to better generalize across diverse problem instances while reducing the need for extensive training data. This approach not only improves performance in out-of-distribution scenarios but also addresses the limitations of existing action-sequence prediction methods. Additionally, the introduction of the Epistemic Planning Domain Definition Language aims to standardize the representation of knowledge and beliefs in multi-agent settings, facilitating interoperability and systematic benchmarking. Concurrently, advancements in goal recognition are addressing biases in existing datasets, enabling more robust evaluations of goal recognizers under varied conditions. These developments collectively suggest a maturation of the field, with a clear trajectory toward more versatile, efficient, and practical planning systems capable of addressing complex real-world challenges.
Top papers
- On Sample-Efficient Generalized Planning via Learned Transition Models(6.0)
- The Epistemic Planning Domain Definition Language: Official Guideline(5.0)
- Removing Planner Bias in Goal Recognition Through Multi-Plan Dataset Generation(5.0)
- PathWise: Planning through World Model for Automated Heuristic Design via Self-Evolving LLMs(3.0)
- Intermediate Results on the Complexity of STRIPS$_{1}^{1}$(2.0)
- Petri Net Relaxation for Infeasibility Explanation and Sequential Task Planning(2.0)