Development of novel approaches to endow robots with cognitive abilities. Ultimately, we want robots to be able to program themselves, given a description of the task in abstract terms.
Examples:
- Deep unsupervised, semi-supervised or self-supervised learning
- Imitation learning / learning from demonstrations
- Sim-to-real and cross-domain transfer learning
- Ontologies and knowledge bases
Advanced Industrial Robotics
Development of novel solutions to challenging problems in industrial robotics, ranging from manipulation to surface treatment.
Examples:
- Efficient and precise path planning in high-dimensional constraint spaces
- Multi-robot collaboration
- Real-time, GPU-accelerated reachability and manipulability analysis
- Real-time multimodal robot and end-effector control
- Precise and data-efficient robot, tool and workspace calibration
Connected Robotics in the Digital Factory
Development of novel approaches for simulating, monitoring and analyzing industrial robots in connected, digital production environments.
Examples:
- Real-time monitoring and anomaly detection for complex robotic production processes
- Prescriptive analytics: Augmenting deep learning with domain knowledge to predict and avoid problems problems before they occur, or deriving solutions to detected problems
- Connected, intuitive user interfaces for programming, monitoring and analyzing robot behavior in the browser, VR and AR
- Edge computing for program analysis and optimization
- Decentralized data storage and federated learning
Medical and Assistance Robotics
Development of novel programming, control and interaction paradigms for assistance robots in the medical and care domains.
Examples:
- High-level, intuitive programming of complex skills for domestic assistance robots by non-experts
- Safe human-robot interaction in the context of elderly care
- Advanced robot control and perception for robotic surgery (such as gallbladder removal)
