Klaus Dorer

About

Klaus Dorer is a leading researcher in humanoid robotics, with a focus on enabling autonomous, real-time performance in dynamic environments. His work centers on computer vision, machine learning, and actuator thermal management for humanoid platforms, particularly in the context of the RoboCup Soccer Adult-Size League. Dorer’s most impactful contribution is the development of a feedforward fully convolutional neural network (FCNN) for the robot Sweaty, which detects and localizes the ball, opponents, and field features in real time. This system, detailed in his 2017 paper (24 citations), can be trained from scratch in just a few hours, marking a significant advance in practical, on-robot vision. He has also pioneered the use of toes in humanoid locomotion and kicking, exploring deep reinforcement learning with PPO to optimize multi-directional kicks in simulated NAO robots (10 citations). Additionally, his 2013 work on evaporative cooling of actuators (9 citations) addresses a critical thermal challenge in high-torque walking and running. Together, these contributions demonstrate Dorer’s commitment to bridging simulation and real-world deployment, advancing both the hardware and software frontiers of autonomous humanoid robotics.

Research Focus

Key Achievements

Top Papers

1. 1
   Detection and Localization of Features on a Soccer Field with Feedforward Fully Convolutional Neural Networks (FCNN) for the Adult-Size Humanoid Robot Sweaty

   24 citations · 2017
2. 2
   Learning to Use Toes in a Humanoid Robot

   13 citations · 2018
3. 3
   Deep Reinforcement Multi-Directional Kick-Learning of a Simulated Robot with Toes

   10 citations · 2021
4. 4
   Evaporative Cooling of Actuators for Humanoid Robots

   9 citations · 2013

Key Collaborators

Contact & Links