Thesis in the Field of Humanoid Robots & AI - Different Focus Topics

At the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB, we turn the latest scientific findings into technical innovations, and in doing so contribute to shaping the future. With your thesis, you will support our Human-AI Interaction team and work on practice-oriented projects carried out in cooperation with industry or research partners. This allows you to apply your theoretical knowledge directly in practice. In our research group, we develop complex experimental setups and demonstrators to test humanoid robots in human-centered environments. We are looking for students to support our team for six months, starting immediately. Just a ten-minute walk from KIT Campus South, your chance awaits to become part of our renowned research institution.

We offer five focus topics. Choose yours

Multi-Agent Orchestration for Humanoid Robots: coordinate and localize multiple humanoid robots Sim-to-Real Transfer for Humanoid Robots: transfer behavior trained in simulation to the real robot Imitation Learning for Humanoid Robots: robots learn fetch-and-carry tasks by imitating human demonstrations Teleoperated Monitoring with a Humanoid Robot: control and monitor a humanoid robot remotely Human-AI Interaction Harness for Humanoid Robots in Care: a runtime environment that allows AI agents to use robots and sensors safely in care settings

Sounds exciting? Then apply now and state in your application which focus topic interests you most. In the interview, we will discuss the topic and direction of your thesis together.

Be part of change

Your tasks depend on the chosen focus topic

• Multi-Agent Orchestration for Humanoid Robots

Research and design of a multi-agent orchestration architecture for humanoid robots in buildings Development of strategies for distributed localization and orientation (e.g., sensor fusion, RTLS integration) Implementation of a prototype for coordinating multiple agents/robots in a building Evaluation of localization accuracy, robustness, and scalability

• Sim-to-Real Transfer for Humanoid Robots

Setup of a simulation environment for tool-fetching tasks (e.g., in Isaac Sim) Implementation of reinforcement learning policies for grasping and transport manipulation in simulation Application of domain randomization and transfer strategies (sim-to-real) Validation of the trained policies on the real humanoid robot and documentation of the transfer

• Imitation Learning for Humanoid Robots

Collection and preparation of demonstration data for tool-fetching tasks (e.g., via teleoperation or kinematic recording) Implementation and training of an imitation learning model to reproduce human fetch-and-carry manipulations Evaluation of generalization to different tools and placement locations Testing on the humanoid robot and documentation of the results

• Teleoperated Monitoring with a Humanoid Robot

Design and implementation of a teleoperation control pipeline for a humanoid robot Development of methods for visualizing sensor and environment data for the operator Testing of low-latency control in human-centered environments (e.g., elevator buttons, doors, corridors) Evaluation of the suitability of teleoperated monitoring in terms of reaction time and situational awareness

• Human-AI Interaction Harness for Humanoid Robots in Care

Design of a generic runtime environment (harness) for AI agents in care scenarios Unified integration of tools and information sources (cameras, sensors, RTLS, HMI systems, robots) Management of context and world state (states of people, objects, and machines) for care activities Implementation of authorization and verification mechanisms for safe action execution Prototypical implementation using the example of "moving hospital beds" and documentation