Reducing Set-Up Times by Designing Grippers in Simulation
In the last 5 years, a method for learning fingers for industrial grippers has been developed in a co-operation between SDU-Robotics and Bialystok Technical University. By this method, cumbersome work on gripper generation, which is a severe obstacle for the fast set-up of robot solutions, could be transferred to simulation.
The figure shows an industrial gripper with fingers learned by the new approach. In this work an objective function for gripper evaluation is derived and then optimized by means of numerical optimization methods. The approach has been evaluated on a variety of real world applications (Wolniakowski et al. 2016) and is currently applied in to a number of industrial use-cases in the ReconCell project. Adam Wolniakowski, who has been supervised by Prof. N. Krüger from SDU-Robotics and Prof. Zdzisław Gosiewski from Białystok Technical University, has defended his PhD successfully on the 10th of November 2016. A description of the approach has just been accepted for a journal paper (Wolniakowski et al. accepted)
A. Wolniakowski, A. Kramberger, A. Gams, D. Chrysostomou, F. Hagelskjær, T. N. Thulesen, L. Kiforenko, A. G. Buch, L. Bodenhagen, H. G. Petersen, O. Madsen, A. Ude and N. Krüger. Optimizing Grippers for Compensating Pose Uncertainties by Dynamic Simulation. IEEE International Conference on Simulation, Modeling, and Programming for Autonomous Robots 2016.
A. Wolniakowski, K. Miatliuk, Z. Gosiewski, L. Bodenhagen, H. G. Petersen, L. C. M.W. Schwartz, J. A. Jørgensen, L.-P. Ellekilde, N. Kruger. Task and Context Sensitive Gripper Design Learning Using Dynamic Grasp Simulation. Accepted for Journal of Intelligent and Robotic Systems.
Full Automatic Workcell Calibration
Calibration between cameras and robots is a difficult problem in many robot vision applications. At University of Southern Denmark we have been busy developing a new method for full automatic workcell calibration. The method we have developed is very user-friendly in the sense that the user is not required to know anything about calibration, cameras and robot control, and the user will be able to calibrate an entire workcell only with the click of a button.
In the ReconCell project the userfriendliness of the calibration method is of great importance, especially when the workcell has been reconfigured to solve a different task, then it might also need to be re-calibrated. The method calibrates the extrinsic parameters of the robot and camera, while it also calibrates the intrinsic parameters of the robot which includes the full Denavit-Hartenberg parameters and Encoder offset.
The calibration method supports multiple types of cameras. You can calibrate both stereo and mono cameras. Multiple types of robots are also supported. Currently we have tested the method with UR5, UR10 and KUKA LBR iiwa. We are now working intensively to extend the calibration method to also support camera-in-hand calibration, where the camera is placed in the robot hand instead of a static placement in the workcell, which is important for quality inspection applications.
Currently the user interface of the calibration method has been implemented in RobWorkStudio (see figure 2), and we are working on integrating a VEROSIM user interface for the calibration method. Furthermore we are working on performing an extensive validation of the method, so far we have performed simple tests with a laser pointer attached to the robot and a checkerboard placed in the workcell. The camera would then detect the four corners of the checkerboard, and the robot would move the laser pointer above the four corners with some known offset. (see figure 1)
White paper on ReconCell workcell design
Several innovations are being developed, implemented and tested as part of the ReconCell project. Reconfigurable fixtures that use a passive Gough-Stewart mechanism enable several different workpieces to be manufactured in the same robotic workcell. The frame of the cell is made of square steel beams connected by special joints. The joints are very stiff, require no welding and are disassemblable. Peripheral application specific elements can be added to the cell using the plug and produce connector. The connector transmits power, Ethernet and pneumatics as well as providing mechanical coupling. An innovative linear unit enables the motion of the robot base. It is unactuated therefore the motion is powered by the robot itself by attaching it’s tip to a fixed point. This significantly reduces costs as no actuators or measuring equipment is needed. The innovative concepts show great potential in making robotic cells modular and reconfigurable. To further confirm the benefits of the proposed innovations more use cases from different industrial backgrounds have to be tested.
ReconCell started its second year
The developments are getting more and more interesting, so this blog will become more active in the following weeks and months. We plan to keep you up-to-date with our latest results and more information about what's new in the project.
Novel Hardware Design
The main results of the first year include a novel reconfigurable hardware design and software architecture for automated robot assembly. We developed new fixtures called hexapods, which are unactuated Gough-Stewart platforms with six degrees of freedom. They have raised great interest at the Automatica 2016 fair in Munich (http://automatica-munich.com/). These fixtures are the building blocks in our novel modular and reconfigurable frame design, which allows easy adaptation of the workcell configuration. Another feature contributing to easy reconfiguration are also the new plug & produce connectors that standardize how a group of modules is connected. Further components that contribute to cheap and automatic reconfiguration are the novel passive linear units that allow us to increase of the robot’s work envelope significantly. We showed in our concept that our workcell can be partially automatically reconfigured to a variety of configuration and use cases.
ROS-based Software Architecture
To drive the passively reconfigurable hardware elements with the robot and carry out a variety of assembly tasks, we developed a ROS-based software architecture that supports communication between all hardware components. It consists of several ROS nodes that implement the basic communication infrastructure and user interfaces. Other nodes can be added based on task requirements, for example “vision module” and “digital interface unit”. A special Simulink-based real-time server was developed to enable efficient robot control. Besides standard industrial trajectory planning technologies, our workcell supports also new advanced trajectory planning techniques based on dynamic movement primitives. These are the basis for learning and adaptation technologies supported by our system, which are essential to increase the number of tasks that can be accomplished without extensive additional programming.
Currently we are applying our system to a number of use cases provided by the industrial partners of the project (ELVEZ, Precizika Metal, LogicData). Stay tuned for more news about exciting new assembly tasks that can be accomplished by the ReconCell.
Here you can see the video from our booth at Automatica, where the robot successfully reconfigures the hexapod and inserts a peg into the hole. This is done using advanced control methods that use the forces that occur during insertion to adapt robot motion.
ReconCell at Automatica fair
ReconCell project to present its results at Automatica fair in Munich, Germany, from June 21st-24th
This year Automatica (Int'l Trade Fair for Automation and Mechatronics) takes place in Munich, Germany, from June 21st-24th. At the fair ReconCell will present some of the key elements of its reconfigurable workcell for fast setup of automated assembly processes. A robot demonstration of an example reconfiguration process that supports fast changeovers of product types will be shown. Besides the robot demonstration, ReconCell will present its design of reconfigurable fixtures, a simulation system for setting up new assembly processes, and an innovative plug-and-produce interface.
European Robotics Forum 2016 Workshop
Efficient robot programming, force control and passive reconfiguration technologies for fast set-up of robotic workcells
Day 3, 23rd March 2016
Rapid changes in market demands lead to decreasing product life cycle times and also more frequent products launches. An enterprise has to react fast, efficiently, and in an economically justified way to market changes. More frequent changeovers in product type or in a number of products require new engineering and production methodologies and machinery equipment to enable shorter set-up times of production environments.
Robots as highly flexible devices have been successfully utilized in many industrial production processes. However, companies are still reluctant to employ robots for many types of tasks. The main hindrances are complexities involved in setting-up robot-based automated solutions because these usually require expert knowledge and also significant time for testing and fine-tuning. Looking at robotic systems in more detail, we can recognise that these problems are due to the long time needed to re-configure and reprogramme the robot workcell for a new production task.
The goal of this workshop is to find out how technologies such as efficient robot programming, sensing, and passively reconfigurable elements can contribute to faster set-up times of robotic workcells. We plan to make the round table discussion available either in video form or as a short report available on the workshop webpage. The results of the workshop will elucidate what technologies are needed to shorten set-up times of robotic workcells and where more research is needed in the future.