Return Flow Handeling

-The team:


Loris van der Voort,  Tim Jansen,  Walter HeemskerkWillem van der Linden

-In collaboration with:



WPS is specialized in smart logistics systems for horticulture purposes. They deliver all kinds of solutions for multiple situations. WPS systems are fully interconnection which means each individual solution is more effective. This means support can be given all the way along the process, from spacing up to delivery. WPS is known for its walking plant systems. In this system the plants move towards the worker and not the other way around.

Current situation

Picture 1

In the current situation WPS delivers transport systems for orchids. The protected growing of orchids is a delicate and well-known process. Currently, these plants are grown in a Plant Growth Cells (PGC’s)(see picture 1) and the PGC’s are transported through the greenhouse as a singular plant on a transport conveyor. Unfortunately, when transported, not all plants are sorted. This results in a return flow of orchids. For the future situation WPS want to explore the possibility of upscaling the input by putting more plants on the same carrier, a tray with six positions. Therefore less automated structure in the greenhouses is needed.

To make the return flow of the orchids more efficient, WPS wants to provide their customers with a solution in the return flow which can detect and sort out the full and empty PGC’s.

Main goals

The goal of the project is to sort out the full PGC’S from the empty PGC’s. With a vision camera, the system needs to detect PGC’s filled with an orchid, an empty PGC’s and empty positions in the tray. Those positions can vary so the robot needs to know what it needs to do.

The PGC’s which contain an orchids has to be put in an oval plant carrier (OPD) See figure 1. OPD’s run over a conveyor which transport the orchids to the next part of the greenhouse. Next up it should stack the empty PGC’s for further use. At last it has to stack all the empty trays also for further use.


  • Vision

The machine vision camera used is called Cognex. This camera can be seen as the eye of the robot, it can detect what positions the PGC’s are in. When it takes a picture the camera takes first looks at the front row of the tray. It recognizes the color black, so when a location is empty it sees a black spot. After it searched for black spots, it will look at the top of the PGC’s. Using a greyscale and threshold the software is able to detect an orchid in a PGC’s with the blob function. If there is an orchid in a PGC it will detect a blob, otherwise it won’t.
After doing this, a string of data is given to the robot. The string exist out of zeros, ones and two’s. Using the data, the robot picks up the three PGC’s and sort them. During this program, the Cognex camera takes another picture of the second row in the tray. It then runs the same program again to sort the second three PGC’s.

  • EOAT

Because handling speed is important for the client, the decision has been made to design an EOAT, which can handle three PGC’s at once. It is designed in a way that it can pick up the first row of PGC’s in a tray. With the same movement of the robot, the two outer ends can spread to the side. By spreading the two outer arms the PGS’s have the same distance as the placement in the OPD’s. This opens the possibility to put three filled PGS’s a once. Furthermore the EOAT is able to pick up the empty trays and move them to the required location. The EOAT is shown in the figures below.




To make the test setup as close to reality, one conveyor is used for the supply of the OPD’s. Normally other conveyors are also used in the system. In this test setup this will be simulated by a light signal.

A locking system is used to make sure, that the OPD’s are locked in a certain position. This allows the robot to place the PGC’s in the OPD’s. Both the locking position and the conveyor are contTestingrolled by the plc. So the system knows when an OPD is filled with an orchid. And sends it though.

To integrate the whole system multiple tests are performed. This went not without any trouble, because the placement in the OPD’s has a low tolerance. It was difficult to get all the positions right. With a lot of trouble shooting the system worked.


Based on the test results, the whole system is able to detect different positions of the PGC’s in the trays. After detection the robot is able to pick up the PGC’s from the tray. According to the different position the robot can sort out the filled and empty PGC’s. It then can place the filled PGC’s in the OPD’s and stack the empty once in a separate tray. Furthermore it can remove an empty tray by picking it up and stacking them elsewhere.

The system can handle a tray including removing the tray itself with a speed of around one tray every 30 seconds.