Focused Model-Learning and Planning for Non-Gaussian Continuous State-Action Systems
This work addresses planning challenges in robotics and control for non-Gaussian systems, though it appears incremental as it builds on existing model-learning and planning methods.
The paper tackles the problem of model learning and planning in stochastic domains with continuous state-action spaces and non-Gaussian transitions by introducing an efficient framework that estimates local models on-demand and focuses on relevant states and actions, achieving asymptotic optimality and demonstrating effectiveness on a simulated multi-modal pushing task.
We introduce a framework for model learning and planning in stochastic domains with continuous state and action spaces and non-Gaussian transition models. It is efficient because (1) local models are estimated only when the planner requires them; (2) the planner focuses on the most relevant states to the current planning problem; and (3) the planner focuses on the most informative and/or high-value actions. Our theoretical analysis shows the validity and asymptotic optimality of the proposed approach. Empirically, we demonstrate the effectiveness of our algorithm on a simulated multi-modal pushing problem.