Venue Details

Electronics manufacturing is often high-mix and low-volume, where programming a robot for a single order can take longer than the job itself. Using the PolyScope X AI Agent Bridge, build an AI agent that turns a customer order into a complete robot program and executes it on a real UR cobot. The goal is a workflow where the robot can move from order to finished electronics product without a human manually programming the cell.

This challenge is co-given by peaq, Doosan and Codecflow. Build a secure, portable robot work plan that can be created in simulation and executed unchanged on a real Doosan A0509. Using peaqOS, teams must constrain the robot to authorized INSPECT, PICK and PLACE actions, recover safely from hidden failures or changed conditions, and generate a verifiable RobotWorkReceipt. The challenge is to prove that a robot can act autonomously while every proposal, decision and result remains policy-compliant, traceable and independently verifiable.

Choose between two manipulation challenges that push robots beyond fixed, pre-programmed behavior. In “Put it Back,” restore a disturbed table or nightstand to a reference state using only a photograph, combining perception, pose estimation, planning and robust grasping. In “Held Between Hands,” develop bimanual manipulation on real hardware for tasks such as pouring, uncapping, insertion, cutting and handover, focusing on data collection, policy design, real-time control and collision-safe execution.

This challenge is co-given by RobCo. Tackle a real production challenge from the printing industry by building an intelligent, vision-driven robotic system for handling print material in a dynamic high-mix environment. The robot should perceive what is in front of it, adapt its movements and recover from unexpected handling failures. Combine sensing, control and decision-making to move beyond fixed programming and create a system that can cope with the variability and unpredictability of real-world industrial production.

This challenge is co-given with TUM MIRMI. Design and prototype an autonomous mobile feed mixing and delivery system for a simulated farm environment. Using a robotic vacuum chassis, build a robot that can identify and navigate between loading and unloading stations, mix feed during transport, determine when the desired mixing state has been reached and dispense it accurately at the destination. The goal is a robust end-to-end system combining autonomous navigation, process optimization and reliable hardware-software integration.

Build an automated pipeline that designs robot grippers for different objects and manipulation tasks. Starting from a description or specification of the parts to be handled, the system should generate a functional parametric gripper design that can be manufactured and tested on a real robot arm. Teams will validate their designs across objects with different properties, from delicate to rigid, with performance ultimately measured by whether the robot can successfully manipulate the predefined parts.

This challenge is co-given with Livekit, and will be held in Garmisch at the TUM MIRMI Garmisch Campus (accommodation is covered for this challenge). Transform GARMI into a natural, multimodal conversation partner for elderly users. Using the LiveKit SDK in the real-time voice pipeline, enable the robot to identify and face speakers, respond fluidly, handle interruptions and natural pauses, and communicate through voice, gaze, head movement, gestures and its chest screen. Explore capabilities such as speaker identification, multi-person conversations, diarization, dialect robustness and privacy-preserving processing, then demonstrate the experience live through a natural, interruptible conversation.

Automate one of the most skill-intensive steps in brick production: building visually balanced shipping bundles from fired bricks with natural color variation. Develop a proof-of-concept that identifies brick or shade categories, plans picks and assembles either smoothly blended bundles or deliberate color mixes. Your solution should combine perception, picking, sequencing and bundling logic while handling textured, non-homogeneous surfaces and preserving the visual appearance expected in the final package and finished wall.