IC4 - TransVision

Localization of partially transparent containers on pallets for robot-based depalletizing

In intralogistics, robots are used for depalletizing packaged beverage containers, for example. The material of the surface plays a major role in detecting the boundaries of the containers to determine gripping points for the robot. Transparent surfaces cannot be detected by standard reflection recognition, because nearly no light is reflected. Therefore, classical approaches to laser measurement technology such as stripe projection do not work.

 

Aim

The research project focused on developing an innovative sensor concept for the localization of semi-transparent packages (e.g., PET bottles) on pallets. The goal was to enable robots to detect these packages in a way that allows for automated and precise depalletizing. The primary challenge was that traditional sensor solutions, such as laser systems, often fail with transparent materials due to a lack of reflection.

 

Solution Approaches

During the project, three approaches were explored to detect transparent packages:

  1. Thermographic Cameras: This technology utilized infrared radiation to make the edges of PET packages visible. However, this approach was not pursued further due to interference caused by the cooling effect of the liquid inside the bottles.
  2. 2D Distance Sensors: This method used ultrasonic sensors, which encountered difficulties in accurately detecting packages positioned side by side.
  3. 3D Reconstruction of Non-Transparent Areas: This approach, which ultimately succeeded, identified non-transparent areas of the packages, such as bottle caps or labels, and reconstructed their positions using an AI-based algorithm.

 

Results

The successful method employed a 3D stripe projection technique that detected the non-transparent areas of the packages. The developed AI algorithm identified the packages based on around 20 real images, supplemented by approximately 1,000 virtual training datasets. This allowed the robot to reliably calculate the gripping points of the bottles with a success rate of 98%.

To validate the results, a demonstrator was built that successfully showcased the functionality in real-world application scenarios. The findings were presented at trade shows and events hosted by the InnovationCampus Future Mobility, garnering significant interest from potential industry partners.

 

Future Prospects

The research outcomes are currently being further developed by PremiumRobotics for integration into commercial applications. A follow-up project, focused on the full detection of transparent areas, is already in the planning stages.

 

Schematic process. © Premium Robotics
Demonstrator setup for evaluating project results. © Premium Robotics

Key data

Research Field

Manufacturing Systems

Period

01.02.2022 until 31.10.2023

Project participants

Contact

Houssem Guissouma

Research Coordinator "Software-System-Architectures"

Phone
+49 172 9830585
E-Mail
fk@icm-bw.de