Overview
Embodied AI, i.e., research combining AI and robotics often requires advanced hardware and systems that are not readily available off-the-shelf. This was especially true in the past, i.e., at the very beginning of the 21st century when research on intelligent robots started under the lead of Andreas Birk at Constructor University (CU). The CU robotics group has hence a long-standing expertise ranging from hardware development to system integration as foundations for research on AI-based robotics, especially related to machine perception.
This started already more than 25 years ago, i.e., at a time before Raspberry Pi or Arduino existed, with research on the CubeSystem, an embedded system framework suited for fast prototyping of intelligent autonomous robots. The CubeSystem and its core hardware, the RoboCube, were also important components for mechatronics developments. One example is the RuggedBot with its special locomotion flipper, which was designed for search and rescue (SAR), respectively general response operations. The flipper of the RuggedBot helps this robot to negotiate difficult terrain and it enables it to, e.g., climb stairs. This flipper is as a special feature actuated by a linear – instead of a standard rotational – drive, which substantially reduces the wear and tear of its mechanical components.
System integration in combination with mechanical engineering also include the development of a deep-sea multi-view stereo-system, which formed the basis of the intelligent machine perception in the DexROV project.