Focus Areas

• Detection, classification, and tracking of arbitrary objects, e.g., clothing items, metal particles, or rock materials.
• Targeted deployment of these objects by autonomous systems such as:
  • humanoid robots,
  • collaborative robotic arms (CoBots), and
  • delta robots.

Another research focus lies in the development of machine learning methods for:

• Sensor fusion from tactile and visual data for robust object detection and interaction in dynamic, unstructured environments,
• Autonomous grasping with robotic hands based on data-driven control strategies,
• Self-supervised skill learning, enabling robots to develop targeted manipulation skills through exploratory learning without external annotation,
• Unsupervised representation learning to acquire compact and generalisable state representations for sensory inputs,
• Reinforcement learning for continuous improvement of motion strategies and interaction capabilities, both in simulations and on real systems,
• Application and adaptation of foundation models, particularly Vision-Language Models (VLMs) and Visual-Language-Action Models (VLAs), for semantic scene analysis, decision support, and planning in robotic applications.

These approaches are combined through hybrid learning strategies and evaluated in real application domains – such as industrial waste sorting, handling and processing of metalworking tools, or surveillance and sorting scenarios with humanoid robots.