Max Robotics — Global Service Robot Directory + US Certification

About

Kenzo Lobos-Tsunekawa is a roboticist at the forefront of integrating deep reinforcement learning (DRL) with autonomous navigation for humanoid and service robots. His research centers on enabling robots to perceive and move through complex environments without pre-built maps, using only visual or point cloud data. His most influential work, “Visual Navigation for Biped Humanoid Robots Using Deep Reinforcement Learning” (101 citations), pioneered a mapless system that extracts motion commands directly from color images using DRL, a significant step toward truly autonomous humanoids. He further advanced sim-to-real transfer in “Point Cloud Based Reinforcement Learning for Visual Navigation” (11 citations), addressing partial observability—a critical challenge in deploying learned policies in the real world. Lobos-Tsunekawa also tackles practical constraints, such as limited onboard computing, as seen in his work on CNN-based object detection for soccer-playing NAO robots (19 citations). His contributions to collision avoidance through multimodal DRL and to accelerating decentralized reinforcement learning demonstrate a consistent focus on making learning-based control both robust and computationally efficient. His work is essential reading for anyone interested in bringing reinforcement learning out of simulation and onto the feet—and wheels—of real robots.