Case Study: DLR achieves 5–10x faster robot training with Google Cloud Platform

DLR Helping robots to help themselves with Google Cloud

DLR’s Institute of Robotics and Mechatronics, part of Germany’s national aerospace institute, wanted to make its robots more intelligent and autonomous using deep learning, but its on-premises servers couldn’t handle the heavy compute demands of training models and running complex simulations. It turned to Google Cloud Platform and Google Compute Engine to support robot learning research, including deep reinforcement learning and physics-based simulations.

Google Cloud Platform implemented scalable Compute Engine virtual machines with flexible CPU and GPU options, allowing DLR to configure powerful instances on demand. The results were significant: DLR says it can train robots five to ten times faster than before, explore cutting-edge research without being limited by computing resources, and improve capabilities such as touch-sensitive material detection, dexterous object manipulation, and symbolic planning.

DLR

Berthold Bäuml

Head of Autonomous Learning Robots Lab