Max Robotics — Global Service Robot Directory + US Certification

About

Ivan Koryakovskiy is a researcher at the forefront of bridging reinforcement learning (RL) and optimal control for real-world robotics. His work centers on a critical challenge: enabling robots to learn complex behaviors without sustaining physical damage during the inherently risky exploration phase of RL. Koryakovskiy’s key contributions include developing methods for model-plant mismatch compensation using RL, which allows robots to adapt to unknown dynamics without requiring extensive, damage-prone trials from scratch. His 2018 paper on this topic has garnered 37 citations, reflecting its significance in the field. He has also systematically benchmarked model-free versus model-based optimal control strategies (28 citations) and rigorously evaluated the physical damage associated with different action selection strategies in RL (7 citations), providing crucial guidelines for safe robotic learning. Beyond RL, Koryakovskiy has applied Nonlinear Model Predictive Control to achieve dynamic squatting motions on the Leo humanoid robot, demonstrating his commitment to real-time, whole-body control. His research is essential for students and engineers seeking to deploy learning-based controllers on fragile, real-world platforms.