ExoPredicator: Learning Abstract Models of Dynamic Worlds for Robot Planning
This addresses the problem of robot planning in dynamic environments with concurrent exogenous processes, representing an incremental advance by combining existing techniques like variational Bayesian inference and LLM proposals.
The paper tackles the challenge of long-horizon embodied planning by proposing a framework that learns abstract world models with symbolic state representations and causal processes for both agent actions and exogenous mechanisms, enabling fast planning that generalizes to held-out tasks with more objects and complex goals, outperforming baselines in five simulated tabletop robotics environments.
Long-horizon embodied planning is challenging because the world does not only change through an agent's actions: exogenous processes (e.g., water heating, dominoes cascading) unfold concurrently with the agent's actions. We propose a framework for abstract world models that jointly learns (i) symbolic state representations and (ii) causal processes for both endogenous actions and exogenous mechanisms. Each causal process models the time course of a stochastic cause-effect relation. We learn these world models from limited data via variational Bayesian inference combined with LLM proposals. Across five simulated tabletop robotics environments, the learned models enable fast planning that generalizes to held-out tasks with more objects and more complex goals, outperforming a range of baselines.