Training Highly Dynamic Robots for Complex Tasks in Industrial Applications

Automation in logistics is moving from automated guided vehicles toward autonomous transport robots. With this shift in robotics, developing logistic systems becomes more and more complex. Fraunhofer IML is known for pushing the boundaries of logistics automation with projects like the highly dynamic LoadRunner, or the recently presented O³dyn pallet transport robot. In this talk, we'll preview what the next generation of autonomous transport robots will look like — agile, flexible, and intelligent. The evoBOT is a bio-inspired assistance robot combining complex skills such as dexterous object manipulation and agile transportation for a wide range of potential industrial applications. Along with this, we'll introduce the concept of guided reinforcement learning, which provides a systematic approach to accelerate training and improve performance in real-world robotics settings by incorporating additional knowledge sources. We’ll show how these technologies, interleaved with Isaac SIM, can pave the way to an industrial application of AI-based robots, both in simulation and reality, and provide deeper insights into the underlying methods used. Check out the following resources below to already get a first preview of what this talk is about. Preview evoBOT robot by clicking this link and Guided Reinforcement Learning by clicking this link .