[2604.08685] RAMP: Hybrid DRL for Online Learning of Numeric Action Models

Computer Science > Artificial Intelligence arXiv:2604.08685 (cs) [Submitted on 9 Apr 2026] Title:RAMP: Hybrid DRL for Online Learning of Numeric Action Models Authors:Yarin Benyamin, Argaman Mordoch, Shahaf S. Shperberg, Roni Stern View a PDF of the paper titled RAMP: Hybrid DRL for Online Learning of Numeric Action Models, by Yarin Benyamin and 3 other authors View PDF HTML (experimental) Abstract:Automated planning algorithms require an action model specifying the preconditions and effects of each action, but obtaining such a model is often hard. Learning action models from observations is feasible, but existing algorithms for numeric domains are offline, requiring expert traces as input. We propose the Reinforcement learning, Action Model learning, and Planning (RAMP) strategy for learning numeric planning action models online via interactions with the environment. RAMP simultaneously trains a Deep Reinforcement Learning (DRL) policy, learns a numeric action model from past interactions, and uses that model to plan future actions when possible. These components form a positive feedback loop: the RL policy gathers data to refine the action model, while the planner generates plans to continue training the RL policy. To facilitate this integration of RL and numeric planning, we developed Numeric PDDLGym, an automated framework for converting numeric planning problems to Gym environments. Experimental results on standard IPC numeric domains show that RAMP significantly outp...