Seite - 26 - in Proceedings of the OAGM&ARW Joint Workshop - Vision, Automation and Robotics

RobWood - Smart Robotics for Wood Industry Thomas Haspl1, Claudio Capovilla1, Alfred Rinnhofer1, Victor J. Exposito Jimenez1, Stefan Maier2, Alexander Heinisch2, Matthias Vo¨lkl3, Manfred Zarnhofer4, Robert A. Jo¨bstl5, Erhard Pretterhofer6, Bernhard Dieber1 and Herwig Zeiner1 Abstract—Many branches of the manufacturing industry in general, and smaller wood processing companies in particular, are facing challenges related to producing ever smaller lot sizes under increasing time pressure. The RobWood project aims to increase the flexibility of such companies by providing a tool-chain to easily program robots for wood processing. In this paper we present an overview on our approach to robot programming by using models of the finished product. I. INTRODUCTION Austrias wood processing industry accounts for 10 billion Euro, and ranks with a 3.9% trade balance surplus on second place just behind the tourism industry (4.2%). Each year, about 18,000 building construction permissions are issued, where prefabricated houses have a share of approx. 30- 35%, with an upward tendency over the last years [2]. Ever higher demands regarding quality standards and individuality pose serious challenges to companies in the wood processing industry. The goal of the RobWood project is to enable strong individualization of products at an equal or higher level of production efficiency through new technological approaches. The integration of robotics, sensor technology, and knowl- edge transfer with appropriate human-computer interfaces, applied inproduction,helps tooptimizeoperatingprocedures in the wood industry. The use cases on which we work on in this project come specifically from the manufacturing of wooden prefabricated houses. Here, every house can be individualized but the parts are prefabricated in a factory instead of building them on site. In order to do this, many different steps have to be performed at each part like cutting, milling and clamping and the joining of different parts like steam brakes with the wooden elements. Model-based programming is a powerful concept, which can lead to more natural interaction and easier programming of industry robots. Employees in smaller wood processing companies without in-depth knowledge regarding traditional robot programming will so be able to program robots them- selves. *The work reported in this article has been supported by the Austrian Research Promotion Agency under grant nr. 849896 1 JOANNEUM RESEARCH {firstname.lastname}@joanneum.at 2 RIB-SAA 3 ABB AG 4 Zarnhofer Holzbau GmbH 5 Haas Fertigbau Holzbauwerk GmbH & CoKG 6 Holzcluster Steiermark GmbH Research into intelligent technologies for accessing the data and knowledge created thereby has a strong leverage effect on its usage, already within single production enter- prises and additionally across company boundaries. The robot based production optimization pursued by the project has enormous potential regarding the creation of new jobs also in more rural areas, the efficient use of resources, and the transfer of insights to other sectors. The rest of this paper is structured as follows: In section II we present related work, in section III we describe the challenges of automated wood processing, in sections IV and V we present the concept and tool chain of our solution and finally we conclude in section VI. II. RELATED WORK The trend towards computer based planning and process- ingmethods hasbeenfinding itsway into the woodmanufac- turing industry sincea fewyears. In somesectorsof theman- ufacturing industry, automated CAD/CAM (computer aided design/manufacturing) systems that generate machining data for use in lot size one manufacturing out of geometrical CAD dataalreadyexist - suchas in theprefabricatedconcreteparts industry, and of course metal cutting on CNC machines as well as additive printing for prototype construction. A. Model-based industrial Robot Programming New research work is investigating new programming methods for making complex tasks easier to program for standard industrial robots [9]. Common approaches include offline programming methods with a complete 3D model [7]. The second common procedure called teach-in, or online programming, is very time consuming for complex task processing. Other approaches such as intuitive robot pro- gramming for SMEs (small and medium-sized enterprises) are described e.g. in [3], [13]. This approach is based on new types of e.g. gesture-based definition of poses, trajectories, and tasks. It is based on a visual programming concept that allows non-skilled programmer operators to create programs. For complex manipulation processes with a huge amount of CAD models, this approach does not significantly reduce the effort for the programming task. B. Model-based Approaches In short, model-driven engineering (MDE) [12], [14] is summarizedas follows:modelonce,generateanywhere.This principle is particularly relevant when it comes to the build- ing of robot applications. The modeling is done on different 26