C06: Lifelong Learning for Autonomous Versatile and Flexible Robotic Grasping System in the Circular Factory

The main goal of C06 is to develop a lifelong grasp learning system for intralogistics handling and manipulation tasks. In collaboration with C03, we aim to deliver fully automated, versatile, and adaptive mobile assistants. To achieve this, a lifelong learning framework is being designed for direct online learning on a real robot, allowing continuous adaptation to variations in object characteristics. To this aim, autonomous exploration-exploitation decision-making based on uncertainty-driven methods is being established to balance learning and task execution effectively. Together with C05, a simulation model will be developed with sim-to-real transfer capabilities. The object detector from B04 will be fine-tuned for dynamic environments on the mobile assistant robot. Smart efficient replay techniques will be devised and achieved with interexchange with INF and C-level subprojects, to overcome catastrophic forgetting in incremental learning as well as to minimize computational costs.