The Need for an Intuitive Teaching Method for Small and Medium Enterprises

Small and medium enterprises currently do not often use robots in their production; one reason is the complex handling, especially the time-consuming programming. Using the well known walk-through approach this drawback could be overcome. We discuss the defiances of this approach and present means of adapting trajectories by transforming them into sequences of geometric primitives. By means to interact these sequences they can get changed according to the specific task. A metric is developed based on triangulation of the trajectories. A gluing robot cell is presented using this technology. Introduction Industrial robots are widely used in companies producing mass market products in high lot sizes. Body shell work in the automotive industry for example mainly consists of robots handling, machining and joining the sheet metal parts. Besides the further development of these applications, industrial robots are currently intruding other markets: in a few years small and medium enterprises will benefit from industrial robots as much as the automotive industry does today. Deficit: Small Lot Size Production In small and medium enterprises robots are not commonly found. One of the reasons, the high investment, is rapidly vanishing: the cost of a robot system has fallen to 25 percent of the costs of 1990 (quality adjusted, [IFR04]). Another reason is the necessary environment, especially the programming capabilities. To work with today’s systems, the SMEs need to set up a robots department with programming engineers and trained service personal. These financial efforts do not pay up till now. In the upcoming part of this contribution, we will focus on the area of programming. Figure 1 : Examples of small lot size processes: welding (left), cutting (right) Definition of Niche Common industrial robots are programmed by a teach panel in lead-through programming, or with an offline programming system. These and other programming possibilities have been described e.g. by Biggs and McDonald [Biggs03]. Both methods only pay for high lot sizes; they need a long time and much experience. Figure 2 shows some connections between lot size, degree of automation and programming method. Offline programming systems will also be used for production lines in the mid or low lot size, single robots will be programmed with the teach panel. In this lot size area human workers can be more efficient than an automated cell or line. Single work pieces will normally be processed by human workers. We propose to get a transit of manual production to automated robot cells for small lot sizes. The main problem, the complex and time consuming teach-in shall be done with the Intuitive Teaching method. Figure 2: Definition of niche Requirements and Applications To succeed with the transit, several prerequisites have to be fulfilled: • New users can get trained to program the robot system in less than one day In small and medium companies the work force is often not big enough to have one person assigned to the robot with a big part of his working time. So long trainings have to be avoided. • Programs can be changed easily When working in small lot size production the number of variants is often high. For work pieces that are nearly the same the same robot program should be used, slightly adapted. • Programming time reduced to a fraction of today’s methods The ratio of programming to production has to be increased for small lot sizes. In the next chapter we propose a programming method that fulfils these conditions. It is especially designed for trajectory oriented tasks because these are the most time consuming. Possible applications are, among others: • Arc Welding • Gluing • Material Handling Proposed Solution: Intuitive Teaching We propose to use a walk-through attempt to provide a tool for fast and effective teaching of industrial robots in this niche. The user guides the robot with a handle that is equipped with a force torque sensor. The robot moves actuated by an admittance control strategy. The trajectory guided by the human is recorded and can be replayed. Before replay, parameters like velocity, position and orientation can be adopted. This programming approach is not new, it has been used e.g. with early painting robots. But today, it is not in use anymore. Our goal is to solve the problems that prevent the usage of this intuitive teaching approach. State of the Art The American Occupational Safety and Health Administration defines three means of programming a robot: lead-through programming, walk-through programming and offline programming [OSHA06]. Today’s robots used in industry rely on lead-through programming with a teach panel, or the offline programming with complex tools. Figure 3: Different means of programming, lead-through (left), walk-through (middle) and offline programming (right). Source: [OSHA06]. The walk-through programming is commonly not used in the industry, but there exist several companies with products in this area. The Barrett arm (Figure 4, left) can be guided by the user, trajectories can be recorded [Leeser94]. Additional functionalities like virtual walls add value. The robot is actuated on the base of motor current measurements. Manutec robots can get equipped with mz robotlab controls, these are able to conduct force sensitive processes like grinding or deburring, also they can be programmed by guidance (Figure 4, right, [Zahn06]). The motion control is done using the measurements of a force torque sensor. KUKA robots can be ordered equipped for safe handling, then the robots can be guided using a 6-DOF-joystick. Figure 4: Barrett WAM [Leeser94], KUKA Safe Handling [Heiligensetzer05], mz robotlab [Zahn06] Defiances The presented robots can be guided by the user, but there is no possibility to overcome the problems that prevent and industrial application up till now. • Precision of path regarding position and orientation: The user cannot guide the robot within a tenth mm or degree, the precision has to be achieved in post processing. • Adaptability of the trajectory: Errors in the teaching process have to be easily overcome, changes should be possible. • Human Machine Interface: The user needs multimodal and intuitive interaction. A few years ago safety was another aspect to prevent applications with direct human robot interaction from being used. But with new technologies like the Reis safety controller or the KUKA safe robot technology, and on the other hand standards like the new ISO10218:2006 allowing the collaboration under defined circumstances, the safety is no problem any more. Human-Machine-Interface The first problem to get robots into SME applications is an intuitive human machine interface. Today’s teach pendants are very efficient, but complex devices. Training takes a too long time for some applications and situations. We propose the combination of several input channels: • Guiding the robot through the trajectory using a force torque sensor • Commanding the robot while guiding using a speech dialog system • Adapting the trajectory by using a PDA and 3-D graphical user interfaces In Figure 5 the different devices are depicted. A common device interfaces allows swapping different devices between different applications. Figure 5: Integration of different devices into the instruction environment, dialogue interface, PDA, 3-D graphical interfaces and force torque sensor. Guidance Up till now the guidance is realized using admittance algorithms; see [Albu-Schaeffer02]. Several means of supporting the user can be added, e.g. the implementation of virtual walls as in [Leeser94]. To avoid unintended motions following a change in orientation the gravitational forces applied by the tool have to be compensated. To do this, the mass and the centre of gravity (COG) have to be known. In the following lines a procedure is depicted how to compute these values with the mounted tool using the robot. The four positions described in Figure 6 have to be reached with the robot. After each position the sensor has to be reset to zero values. After reaching position two, three and four relative values of forces and torques are measured. Figure 6: Positions to compute mass and COG From these measurements mass and centre of gravity can be computed: ) ( 3 1 3 2 1 y z x dF dF dF m + + = ; ⎪ ⎩ ⎪ ⎨ ⎧