IKV Logo Portrait

Student paper

Development of an AI-controlled low-cost robotic cell for post-processing in additive manufacturing

Bachelor’s Thesis or Master’s Thesis

Teach low-cost robots to see! Train VLA models for autonomous part removal & revolutionize post-processing in 3D printing
Stuwi-Entwicklung einer KI-gesteuerten Low-Cost-Roboterzelle für das Post-Processing in der Additiven Fertigung© IKV
Figure 1: Automated 3D printing cell – robotic arm and printer operating in sync. Image generated by AI.

Topic of the work:

Additive manufacturing offers maximum design freedom, but post-processing (e.g. component removal) is often still expensive manual work. In this work, we develop hybrid automation concepts for low-cost robot arms that combine classical, deterministic programming with state-of-the-art AI. To maximize traceability and process reliability, we use system and CAD data (STEP/STL/G code) to mathematically calculate safe macro movements and optimum lever arms in advance. However, since inexpensive robots have mechanical inaccuracies and no force sensors, purely deterministic control systems reach their limits when gripping. This is where we come in: We teach the robot the final, contact-intensive fine manipulation via “imitation learning” (Hugging Face LeRobot). A state-of-the-art vision-language-action model (SmolVLA) learns to compensate for the inaccuracies purely visually. The work thus combines the reliability of classic design and control methods with the adaptive flexibility of data-driven control approaches.

The work is related to this research project:

The work takes place in the Additive Manufacturing working group and is directly embedded in the Cluster of Excellence “Internet of Production” (IoP). At the IoP, we are researching how the production of the future can be made more flexible and efficient through the consistent networking of data and the use of machine learning.

Objective:

The aim of the work is to design and commission a camera-supported low-cost robot cell (SO-ARM100) for the automation of handling tasks in 3D printing. Based on this, suitable gripping strategies are derived from CAD data, real training data is recorded via teleoperation and a vision-language-action (VLA) model for intelligent robot control is trained.

Your task:

The work is hands-on and includes setting up hardware, recording robot movements and training and evaluating neural networks on our workstations.

Concept research: Analysis of automation potential in post-processing (e.g. removal by extruder vs. direct gripping) Concept & design: Development of a specialized end effector for the mechanical removal of support structures
Hardware setup: Production and assembly of the robot arms and integration into the LeRobot framework incl. Camera system KI modeling: Training of the VLA model to differentiate between component geometry and support structure as well as for path planning for destructive movements.
CAD analysis & strategy: Use of STEP/STL data to identify optimal gripping points and lever arms for selected components Validation: Benchmarking the success rate of automated decoring of complex geometries compared to manual processing
Data collection: Conducting teleoperation demonstrations for a basic pick-and-place task of AM components Evaluation & benchmarking: Systematic comparison of the robustness of the VLA model under changing lighting conditions or new component colors

Your profile:

  • Basic programming skills in Python are a prerequisite.
  • Initial experience in CAD, 3D printing or machine learning (e.g. PyTorch) is a big plus, but not a must.
  • Interest in finding pragmatic solutions, a “maker” mentality and a desire to work at the interface of plastics technology and artificial intelligence

Preferred study programs:

  • General mechanical engineering
  • Industrial engineering, specialising in mechanical engineering
  • Automation technology
  • Computational Engineering Science (CES)
  • Simulation Sciences

These are your benefits:

  • Work in a young, motivated team
  • Independent work with intensive support
  • Co-design of current research projects
  • Individual coordination of tasks and time frame
  • Experimental work on industrial plants
  • Fast processing, direct contact for questions during processing
  • Varied mix of research, experimental work and analytics (design, AI training, 3D printing, robotics)
  • Immediate start possible

If you are interested in writing your thesis at IKV and in this task, please get in touch with me. We agree on the exact scope of content and schedule individually.