Two papers by our PhD students involved in the ERC project STEP-RL have been accepted at the 28th European Conference on Artificial Intelligence (ECAI)!
1️. "Exploiting Symbolic Heuristics for the Synthesis of Domain-Specific Temporal Planning Guidance using Reinforcement Learning" by Irene Brugnara, Alessandro Valentini and Andrea Micheli:
In this work, we propose a novel framework based on reinforcement learning to synthesize domain-dependent temporal planning guidance. We explore different ways to exploit symbolic heuristics during both the training phase and the planning phase. At training time, we learn a residual of a symbolic heuristic instead of learning the whole heuristics from scratch; at planning time, we combine the learnt information with a symbolic heuristic in a multiple-queue approach.
See the pre-print at https://arxiv.org/abs/2505.13372
2️. "Learning of Lifted Macro-Events for Heuristic-Search Temporal Planning" by Alessandro La Farciola, Alessandro Valentini and Andrea Micheli:
In this paper, we generalize macro-events (introduced in our previous work [1]) to a lifted representation, making them independent of specific problem objects. We devise a fully-automated framework that, given a domain and a collection of small training problems, constructs and selects a suitable set of lifted macro-events. We define a learning pipeline that mixes the optimization of the statistical expectation on an abstraction of the problem with an empirical refinement of the selection on a validation set.
[1] La Farciola Alessandro, Valentini Alessandro, and Micheli Andrea. (2025). Automatic Selection of Macro-Events for Heuristic-Search Temporal Planning. AAAI 2025. https://ojs.aaai.org/index.php/AAAI/article/view/34859