Papers
5
Total Citations
79
H-Index
3
About
Ostap Okhrin is a researcher at the forefront of applied reinforcement learning (RL), with a particular focus on autonomous systems, dynamic obstacle avoidance, and car-following behavior. His work bridges the gap between theoretical RL methodology and real-world deployment, leveraging sophisticated simulation environments — most notably the CARLA driving simulator — to develop and validate intelligent decision-making agents. Okhrin's most cited contribution, a modified Deep Deterministic Policy Gradient (DDPG) car-following model trained on real-world human driving data, has garnered over 40 citations since 2022, reflecting significant community interest in human-informed autonomous driving. Complementing this, his research on collision risk assessment for dynamic obstacle avoidance tackles a fundamental challenge in RL training: the rarity of critical edge-case experiences. By rebalancing encounter probabilities during training, his methods produce more robust and safety-conscious agents for mobile robots, ships, and drones. More recently, Okhrin has turned his attention to overestimation bias in value-based RL, proposing a novel two-sample testing framework to mitigate maximization bias — a subtle but consequential flaw that can cause algorithmic failure. Across his body of work, Okhrin demonstrates a rigorous commitment to making reinforcement learning safer, more reliable, and practically deployable in complex real-world environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5