Max Robotics — Global Service Robot Directory + US Certification

About

Frank Sehnke is a researcher whose work spans reinforcement learning, evolutionary computation, and robotics, with particular expertise in policy gradient methods and parameter-based exploration. He is best known for pioneering **parameter-exploring policy gradients (PEPG)**, a family of reinforcement learning algorithms that directly perturb the parameters of a policy's function approximator rather than injecting noise into actions. This approach elegantly bridges reinforcement learning and black-box optimization, offering more stable and efficient learning in high-dimensional continuous control tasks. His landmark 2009 paper on this topic has accumulated over 245 citations, establishing it as a foundational reference in the field. Alongside collaborators, Sehnke further developed and analyzed these methods across multiple publications, demonstrating their practical applicability to robotic control problems. Earlier in his career, he contributed to the RoboCup robotics domain, developing automatic calibration and online color training systems for mobile robots using evolutionary algorithms — work that addressed real-world sensor calibration challenges without manual intervention. Collectively, his research has meaningfully advanced the theory and practice of autonomous learning systems, making his parameter-exploration framework a lasting contribution to the reinforcement learning community.