Software- und Scientific Team-Projects

The assembly station is a new, stationary robot arm introduced with the Smart Manufacturing League. It specializes in stacking raw material blocks into finished products and, in the long term, recycling these products back into individual blocks.

An Igus Rebel cobot for this task is already available, but a ros2_control hardware integration is still needed.

The robot transports objects within its inventories. With the Smart Manufacturing League, the long-standing limitation of a maximum of three objects that could be transported on the robot, which had defined the @work League, is eliminated. The types of objects to be transported also change. These two factors together enable a fundamental redesign of the robot's inventories.

This project presents the development of a robust grip detection system for the fingers of our robotic. The goal is to reliably detect, using sensor data (e.g., force, pressure, or touch sensors), whether and how securely an object has been gripped. Suitable algorithms will be developed and compared, including threshold-based methods and machine learning approaches. A key focus is on robust real-time evaluation and integration into the ROS-2 robotics frameworks.

Comparative study of integrating a microcontroller to control a gripper servo via micro-ROS and a serial interface. The aim is to establish reliable communication between a microcontroller (Raspberry Pi Pico 2) and a computer and software stack based on ROS 2. Two signal transmission approaches will be implemented and compared. A key focus is on the robust implementation of position and force control, as well as the feedback of sensor data.

This paper presents the configuration and optimization of a navigation system based on the Nav2 stack for the KUKA youBot mobile robot. The goal is to enable reliable autonomous navigation indoors, including localization, path planning, and obstacle avoidance. Key components such as AMCL, Costmaps, and controller plugins will be tailored to the specific characteristics of the robot.

Development of a system for object recognition and pose estimation of SML interlocking building blocks: The goal is to reliably identify individual building blocks using RGB-D camera data and determine their 6D pose in space. To achieve this, modern deep learning approaches will be compared and trained for our specific use case. A key focus is on robustness under varying lighting conditions and partial occlusion.

Development of a system that flexibly identifies and classifies the assembled form/product of the Smart Manufacturing League based on recognized building blocks and their poses. The goal is to design a configurable pipeline that compares the arrangement of individual blocks with predefined models. Graph-based or rule-based approaches as well as machine learning methods will be investigated. A particular focus is on extensibility, allowing new products to be added and old ones removed without significant effort.

This project presents the development of an inventory vision system for verifying the successful gripping and correct placement of building blocks in the storage area. The goal is to automatically detect, using camera data, whether an object has been successfully gripped and correctly placed in its intended position. In case of incorrect grips or placement, the system should trigger a new picking attempt. Robust image processing methods are employed to reliably detect deviations from the intended target state.

Development of a time-saving process plan for object search, grasping, and placement: The goal is to significantly reduce the overall cycle time of a robotic system through intelligent planning and coordination. Strategies such as parallel processing, optimized motion planning, and a state-machine-based approach will be compared and contrasted with a behavior tree. Particular focus will be placed on dynamic adaptation to changing environmental conditions and uncertain perceptual data.