Safe non-smooth black-box optimization with application to policy search

Abstract

For safety-critical black-box optimization tasks, observations of the constraints and the objective are often noisy and available only for the feasible points. We propose an approach based on log barriers to find a local solution of a non-convex non-smooth black-box optimization problem $\min f^0(x)$ subject to $f^i(x)\le 0,i=1,\dots ,m$, at the same time, guaranteeing constraint satisfaction while learning an optimal solution with high probability. Our proposed algorithm exploits noisy observations to iteratively improve on an initial safe point until convergence. We derive the convergence rate and prove safety of our algorithm. We demonstrate its performance in an application to an iterative control design problem.

References in corpus (5)

• Global Convergence of Policy Gradient Methods for the Linear Quadratic Regulator
• The Complexity of Large-scale Convex Programming under a Linear Optimization Oracle
• Model-Free Linear Quadratic Control via Reduction to Expert Prediction
• Worst-case iteration bounds for log barrier methods on problems with nonconvex constraints
• Log Barriers for Safe Non-convex Black-box Optimization