~~NOTOC~~

====== Projekt: ACRAMbly ======

|< 100% 25% 75%>|
^ECTS^12^
^Art^Masterprojekt^
^Semester^SoSe 2025 & WiSe 2025/2026^
^Vortragende^Prof. Michael Beetz^
^Übungsleitung^Jonas Dech, Tom Schierenbeck, Malte Huerkamp, Abdelrhman Bassiouny, Max Gandyra, Duc Than^
^Sprache^ Englisch^
^Plenum^Freitag: 9:30 - 10:30, Ort: TAB (Ecotec 5), Eingang E, Ebene 0 ^

===== Short description =====

In the industry robots are employed for a long time to ease the load of human workers. They are
mainly doing things like lifting and moving heavy parts, for example, in the assembly lines of cars or
welding these parts. However, these robots are only performing the same task over and over again
without being aware of what they are doing or being able to adjust to changes. Furthermore, are
these robots often programmed by hand for a specific set of motions
ACRAMbly (PyCRAM + Assembly) combines the fields of industrial autonomous production with
the field of cognitive robot control. Cognitive robot control addresses the aforementioned problem of
robot control by incorporating artificial intelligence approaches with robotics and enabling robots to
be aware what the actions they are performing are meant to achieve and how they are related.
Specifically, ACRAMbly is about the construction of an automated assembly line using a cognitive
architecture. This involves several steps which will be pursued in this project. First the robot
assembly station needs to be set up and integrated with the control, motion and perception
frameworks. For detecting the accurate poses of the parts in the assembly area we will train a multi-
step perception system. Grasping and assembling is done using two UR10 robot arms equipped with
Robotiq gripper and a control framework. Lastly, all these components are controlled by the
cognitive architecture PyCRAM which sits at the heart of the whole software stack and is
responsible for orchestrating the whole process as well as sending and receiving information to the
other frameworks.
As parts for assembly we will start with simple cubes which need to be stacked and arranged in
order followed by a toy model plane which requires more precision in the movement of the robot
arms as well as the right order of assembly steps.

There are different goals for this project:
   * Integrate an assembly station into the cognitive architecture
   * Plan a sequence of actions which result in a fully assembled plane
   * Precisely estimate the 6D Poses of objects in the assembly area
   * Control the robot arms with high precision to facilitate the assembly process 

===== Recommended Lectures =====

Dies ist eine kontinuierlich wachsende Liste mit komplementären Lehrveranstaltungen, die wir empfehlen:
  * Lecture Course: Grundlagen der Künstlichen Intelligenz (SoSe)
  * Lecture: [[:teaching:suturo-crashcourse_ss25|KI-basierte Robotersteuerung (KIBA)]]