Research map / Reinforcement learning / Policy gradient and actor-critic
Safe-PG-LQR: Provably Safe, Convergent and Optimal Model-Free Linear Quadratic Regulator with Hard Constraints
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes Safe-PG-LQR, a model-free policy gradient method for the LQR that enforces hard state/input constraints throughout learning while retaining convergence and optimality guarantees.
Problem
Model-free LQR control with hard constraints enforced throughout learning; existing methods are partially model-based, use soft constraints via Lagrangian relaxation, or rely on projection.
Approach
A model-free policy gradient algorithm for LQR that keeps every iterate feasible with respect to hard constraints, rather than relying on Lagrangian relaxation or projection, while preserving global convergence to the optimal controller.
Limitations
- Full text unavailable; limitations not stated
Experimental setup
Baselines: Lagrangian relaxation methods projection-based methods partially model-based methods
Builds on: policy gradient LQR global convergence
Closest papers by meaning
- Online Policy Gradient for Model Free Learning of Linear Quadratic Regulators with $\sqrt$T Regret (ICML 2021 · similarity 0.90)
- Efficient Optimistic Exploration in Linear-Quadratic Regulators via Lagrangian Relaxation (ICML 2020 · similarity 0.89)
- DiLQR: Differentiable Iterative Linear Quadratic Regulator via Implicit Differentiation (ICML 2025 · similarity 0.86)
- Regret Bounds for Episodic Risk-Sensitive Linear Quadratic Regulator (ICLR 2025 · similarity 0.86)
- Handling Heterogeneous Curvatures in Bandit LQR Control (ICML 2024 · similarity 0.85)
- Constrained Optimization From a Control Perspective via Feedback Linearization (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.