How REVA University Innovated Automated Gluing With a Cobot

Automatic gluing involves using a robot to apply adhesive along a programmed path. Students at REVA University created a collaborative robotic system using an innovative vision algorithm to reliably glue parts regardless of their placement.

This research study presents some valuable lessons if you are also considering automating a gluing process using robotics. By incorporating cutting-edge AI-based part detection, the project opens up a whole new level of flexibility for this common manufacturing task.

Here’s how the research team at REVA University created this automated gluing system with the help of RoboDK.

Introducing… REVA University Robotics Team

This exciting project comes from a small team of researchers at REVA University, based in Bengaluru, India. Its Faculty of Engineering Technology offers various programs specializing in artificial intelligence and machine learning, as was a focus of the project.

The researchers shared their results with the wider robotic community in a recent IEEE conference on Digital Signal Processing, headed by Prof. Bharath S.H, Akshay Ajit Bhawar, and Joshua John from the School of Electronics and Communication Engineering, whose specialisms include signal processing and AI robotics.

What Is the Gluing Process and Why Automate It?

The team at REVA University chose one of the most valuable tasks to automate in many manufacturing processes: robotic gluing.

Gluing involves joining parts with an adhesive such as hot melt glue, cold glue, or epoxy. It’s one of the 17 most common industrial robot applications and is an ideal task for robotic automation as it guarantees high adhesive consistency.

With manual gluing, high consistency is extremely difficult to achieve. It’s almost impossible for a human to keep a uniform bead size over the course of a long shift. This leads to defective gluing patterns, higher scrap, and a higher cost per part.

Three major benefits of automating your gluing task with a robot include:

  1. High Consistency — An automated system dispenses the same amount of adhesive every time, leading to uniform quality from the first part of the day to the last.
  2. Less Waste — Accurate dispensing cuts reduce the number of faulty products generated by human error, improving the sustainability of the entire process.
  3. Safety for Workers — Using a robot reduces contact with solvent fumes, awkward physical postures, and repetitive physical tasks for your skilled human workers.

Combining these with collaborative robots means that the robot can operate around human workers, reducing the need for safety fixtures.

How Did the REVA University Team Build an AI-Guided Gluing Cobot?

The project described in the IEEE paper started with a long-standing industrial problem statement from the academic literature. The challenge was this: Using point cloud scanning for detecting workpieces was overly complex due to instability, high costs, and a dependence on complex laser cameras. The use of ROS-based programming, which is common in research, also contributed to this challenge.

To solve this problem, the research team proposed using a solution incorporating RoboDK, Python, and a feature-based vision process. This novel approach was the first that integrated a type of part detection algorithm called YOLO and incorporated it with RoboDK’s gluing features.

Vision: Finding the Part Without Markers with YOLO

A core contribution of this research was the use of the YOLOv8 model for part detection. The team used an Intel RealSense camera to capture images and detect the part for gluing through a model trained on custom-labeled images.

YOLO (which stands for You Only Look Once) is an advanced computer vision module with an open-source codebase. It uses AI neural networks for a range of vision tasks, including detection, classification, and segmentation.

The researchers curated a custom dataset of component images and annotated them for features of interest, including circular openings, cavities, and reference holes. They trained these machine learning modules using GPU acceleration and chose a YOLO-based object detection model that could operate in real time.

RoboDK: Turning Detection Into a Reliable Gluing Path

An important contribution of the team’s research is their choice to use RoboDK. This set the research apart from others in the research space, which often use a ROS-based middlewhere.

RoboDK offered the team a more reliable tool for robot programming as ROS can introduce unnecessary system complexity. Previous projects of this type found that ROS can result in unstable performance for continuous industrial use.

In this project, the team used RoboDK in this manner:

  1. Output from the Intel RealSense camera was passed through the real-time YOLOv8 model. This resulted in detection of the full 360° orientation of the part.
  2. This was output to a 4×4 transformation matrix, which applied to the reference frame and glue curve in RoboDK. To do this, the team used the powerful RoboDK Python API.
  3. The resulting glue path was stored as a curve-follow project within RoboDK.
  4. RoboDK post-processors then turned these paths into joint commands for the team’s chosen robot, the Universal Robots UR3
  5. These were then streamed in real time to the UR3 over an Ethernet connection.

The output of this operation was a consistent glue path using a pneumatic syringe with a 0.3 mm nozzle. The dispenser was designed for use with a foot pedal, so the team wired the pedal to a robotic digital output, which was activated through the RoboDK program.

[ROBODK SCREEN]

What Results Has the Team Achieved, and What Comes Next?

The output of this project showed that feature-based detection and RoboDK were a reliable solution for flexible gluing that doesn’t rely on external markers or costly scanners.

The research team’s ingenious system achieved robust part alignment, trajectory adaptation, and synchronized glue deposition in real time with minimal error rates and deviation.

This project is just the beginning. The team’s future plans include:

  • Scaling the system to multi-part assemblies.
  • Integrating advanced deep learning for complex part geometries.
  • Extending the application to processes like sealing, welding, and coating.

The team hopes that these further advances will allow the robot to approach workpieces from different angles and enable adhesive deposition on more complex geometries while maintaining precise tool alignment and trajectory control.

We look forward to seeing how this team of researchers incorporates RoboDK even further in their future projects!

What gluing application do you think would benefit from a robot? Join the discussion on LinkedIn, Twitter, Facebook, Instagram, or in the RoboDK Forum.. Also, check out our extensive video collection and subscribe to the RoboDK YouTube Channel

About Alex Owen-Hill

Alex Owen-Hill is a freelance writer and public speaker who blogs about a large range of topics, including science, presentation skills at CreateClarifyArticulate.com, storytelling and (of course) robotics. He completed a PhD in Telerobotics from Universidad Politecnica de Madrid as part of the PURESAFE project, in collaboration with CERN. As a recovering academic, he maintains a firm foot in the robotics world by blogging about industrial robotics.

View all posts by Alex Owen-Hill →
Scroll to Top