In the robot learning lab, we work on various topics at the intersection of manipulation, robot vision, machine learning, and control. We aim to develop methods and algorithms for learning vision-based manipulation tasks, system identification, reinforcement learning, and optimal control. We work both in simulated and real-world environments to train our agents. We are working on a common platform/benchmark for perception-based manipulation that could be used by the manipulation (learning) community to compare and evaluate the latest advancements. For physical setups, we currently have a single-arm reference setup, with the needed software to interface with it using c++ and python. We also have the required tools (calibration and communication) to quickly mount a bimanual setup whenever a second robot is available.