Complex use cases require sophisticated know how
The automation problems we tackle have one thing in common: They are enormously complex, require non-repetitive motion patterns and deal with dynamic process parameters. Some use cases even demand multiple robots to work simultaneously in the same workspace. This complexity demands advanced robotic solutions that exceed the state of the art. sewts' robotics team develops these technologies in-house. Our core competencies include on-the-fly motion and trajectory planning, real-time control of various industrial robots as well as multi robot coordination.
This is where the real magic happens. Smart algorithms are needed to build adaptive systems that cope with non-deterministic automation processes. That’s why we leverage the latest AI research findings, refine them for our needs and finally put them together to one big piece – our robotic brain. It receives diverse sensor data (e.g. optical information), draws conclusions on a human-like level of cognition and translates these into high-level robotic commands. That’s how our systems complete tasks that until now required human intellect.
Visual information is the most important input for our robotic brain and therefore it is crucial to always provide high quality data. Depending on the use case we apply state of the art 2D or 3D vision systems – both seamlessly integrated in our systems. We are experts in enhancing the generated data which is especially relevant when working with 3D point clouds. This preprocessing is a vital building block of our systems in terms of generating usable input for our artificial intelligence.
The majority of our use cases deal with flexible materials, textiles and alike. It is essential to understand the characteristics of these materials in order to implement robust processes. We achieve that through highly sophisticated material simulations. We develop specialized FE simulations in cooperation with TUM to reproduce the behavior of textiles. These simulations open up completely new possibilities in the development of smart algorithms. We use these simulations to generate synthetic training data for our AI engineering to name just one example – leading to dramatically shortening our development cycles.