Wednesday, April 5, 2017

CATCH @European Robotics Forum 2017 in Edinburgh


The CATCH project represented by Gehard Schreck (IPK) took part in a workshop session with the title "Agri-Food - Robotics for Agri-Food: Echord++ Experience" on this years European Robotic Forum (ERF) in Edinburgh.

Presentation slides: Link

Below you will also find an article extraction (in german, english translation below) from the website heise-online.de summarizing the workshop session at ERF 2017.

Article from 24.03.2017 found at heise.de, Source: Link

"Automatisierung der Gurkenernte

Zum Beispiel bei der Gurkenernte, deren Automatisierungspotenzial im EU-Projekt CATCH untersucht wird. Die derzeit vorherrschende Erntemethode besteht darin, dass bis zu 40 Arbeiter bäuchlings auf einem Fahrzeug liegen, das langsam übers Gemüsefeld fährt, während sie die Früchte pflücken. Bis zu 13 Gurken pro Minute schaffe ein Erntehelfer, sagte Gerhard Schreck vom Fraunhofer-Institut für Produktionsanlagen und Konstruktionstechnik (IPK). Die Einführung des Mindestlohns in Deutschland habe aber die Kosten pro Frucht erhöht und damit auch den Druck in Richtung Automatisierung. Die soll mithilfe von Roboterarmen geschehen, die auf dem Erntefahrzeug montiert sind.

Eine große Herausforderung stelle die Erkennung der Gurken dar, die teilweise von Blättern verdeckt sein können, deren Farbe zudem ähnlich ist. Veränderliche Lichtverhältnisse machen die Aufgabe nicht einfacher, ebenso die unregelmäßigen, völlig zufällig gestreuten Positionen, in denen die Früchte sich befinden. Multispektralkameras sollen helfen, die Arme an die richtige Stelle zu dirigieren. Was für Greifer sich an deren Spitze befinden werden, sei noch nicht entschieden, so Schreck. Derzeit seien drei Varianten in der Diskussion. Ein Feldtest des Systems ist für den kommenden Juli geplant. Es sei eine Besonderheit der Agrar-Robotik, betonte Schreck, dass man sich die Termine für solche Tests nicht frei aussuchen könne, sondern sich an den Reifezyklen der Früchte orientieren müsse."


english translation:

Automation of cucumber harvesting

For example in the case of cucumber harvesting, whose potential for automation is investigated in the EU project CATCH. The currently prevalent harvesting method is that up to 40 workers are lying on a vehicle slowly moving over the vegetable field while picking the fruits. Up to 13 cucumbers per minute can be produced by a seasonal worker, said Gerhard Schreck of the Fraunhofer Institute for Production Systems and Design Technology (IPK). However, the introduction of the minimum wage in Germany has increased the costs per fruit and thus the pressure towards automation. This is to be done using robotic arms mounted on the harvesting vehicle.

A major challenge is the detection of the cucumbers, which can be partially obscured by leaves, the color of which is similar. Variable light conditions do not make the task easier, as are the irregular, completely randomly scattered positions in which the fruits are located. Multi- spectral cameras should help to direct the arms to the right place. What grippers will be at their head is not yet decided, Schreck said. Three variants are currently under discussion. A field test of the system is planned for next July. It was a special feature of agricultural robotics, stressed Schreck that the dates for such tests could not be freely chosen, but must be oriented to the maturity cycles of the fruits

Work@CATCH project



Tuesday, April 4, 2017

CATCH Workshop Review

The CATCH Workshop I held on 6.12.2016 at ATB in Potsdam with representative cucumber cropper and producers from predominant agricultural areas in Germany (Bayern and Brandenburg) have contributed to critically evaluate the initial concepts and modules design, as well as initial testing plan.

The following points relevant to the system and modules concepts and experiment plans have been considered:

  • The adopted low-priced concepts, based on reconfigurable and modular technology are very suitable from the end-user view (the costs of the final systems should be between 10-20 T€);
  • Cucumber-flyers concepts should also be applicable for the robotic systems;
  • Estimated goal productivity estimation 10-12 cucumbers/ min is quite promising and applicable;
  • The preliminary localization results require further improvements in order to reach faster detection at higher rates;
  • The differentiation of cucumber dimension (avoiding the picking the fruits under some limit sizes) should also be applied;
  • The best CLIP action algorithm for robotic applications is still not clear and should be further investigated. The best manual practice approaches from different location demonstrate different strategies. Avoiding fruit and stem injuries is relevant to minimize damages by released cucumber enzymes.
  • By the genetic research and developments the crops can be influenced to provide more faster and robust fruit detection and localization (e.g. by reducing the leaves density and stem characteristics)

As a conclusion a big interest of all end-users for the CATCH robotic technology and experiments, identified as feasible and promising approach for solving critical problems related to the high labour costs and availability of manual workers, as well as ergonomic safety, has been expressed. All participants have mentioned interest to actively follow and contribute by experience CATCH developments using CATCH-blog and planned presentations and meeting according to the elaborate experiments plan.


Presentations:

Maurice Gohlke (Bosch Deepfield Robotics) - 

Dragoljub Surdilovic (Fraunhofer IPK) - PDF (1), PDF (2)

Roemi Fernández (CISC) - PDF

Jelena Surdilovic (ATB Potsdam) - PDF