Get Started
Home
Topics
Search
Library
Research questionHow can we learn an effective LQR controller from unknown dynamics without a stable initial policy?Unknown system dynamics make it difficult to learn a controller with limited interaction data. Requiring a stabilizing initial policy can also exclude systems where no such controller is available beforehand.
AI
Machine Learning
Reinforcement Learning
Research Paper
Statistical Machine Learning
Technology
Latest papersRecent research connected to this question, newest first.Sample Complexity of Linear Quadratic Regulator Without Initial StabilityApplies to linear-quadratic regulator problems with unknown dynamics and potentially unstable initial policies. The source discusses avoiding two-point gradient estimates, analyzing Riccati-operator error contraction under Riemannian distance, and improving convergence and sample-complexity guarantees; it does not provide details about particular system dimensions or deployment settings.research paper · Sep 2, 2026
Related questions
How can model predictive control remain feasible and stable for nonlinear systems with intermittent state measurements?How can model-free robust Q-learning learn from one trajectory with second-moment targets and a noncontractive Bellman operator?How can unknown nonlinear ODE dynamics and time-varying external forces be learned from one trajectory without physics priors?How can multilayer RNNs learn nonlinear quadrotor dynamics without vanishing or exploding gradients?