Ilnura Usmanova

About

Ilnura Usmanova’s research lies at the intersection of safe black-box optimization and reinforcement learning, with a focus on developing algorithms that can operate reliably in safety-critical, real-world environments. Her major contributions center on the use of log barrier methods to handle unknown constraints during online optimization—a challenge common in manufacturing, robotics, and autonomous systems where evaluating an objective requires experiments on deployed hardware. In her 2022 paper, “Log Barriers for Safe Black-box Optimization with Application to Safe Reinforcement Learning” (4 citations), she introduced a framework that enables safe exploration by leveraging noisy feedback on constraint proximity, ensuring that learning does not compromise system integrity. Her earlier 2019 work extended these ideas to non-convex settings, addressing smooth optimization over compact domains defined by unknown functional constraints. Though early in her career, Usmanova’s work is notable for bridging theoretical rigor with practical safety guarantees, offering a principled path toward adaptive control in high-stakes applications. Her research is particularly relevant for students and engineers seeking to deploy learning systems where failure is not an option.

Research Focus

Key Achievements

Top Papers

1. 1
   Log Barriers for Safe Black-box Optimization with Application to Safe Reinforcement Learning

   4 citations · 2022
2. 2
   Log Barriers for Safe Non-convex Black-box Optimization

   2 citations · 2019

Key Collaborators

Contact & Links