Page - 78 - in Proceedings of the OAGM&ARW Joint Workshop - Vision, Automation and Robotics

iterations on the test system and the point cloudmodel can not be matched. The low accuracy of the ICP algorithm in the experiments for lowspeedsmayoccur to theverybumpy surface images from the cameras (Figure 1), whichmakes it difficult to calculate an accuratematch. A smoothing of the robot’s point cloudby themoving least squaresmethod from thePCLalsodoesnot significantly improve the results, since the outliers’ in the robot’s point cloud surface are too large (cf. Figure 2c) to achieve good results. VI. CONCLUSION&OUTLOOK A robot point cloud model generated from CAD data for each robot link have been adopted and linked via the DH convention. A linked motion algorithm is integrated so that each link depends from each other. The first tests with structured light cameras and the ICP algorithm from the PCL showed moderate results. For the next tests with structured light cameras, the results should be improved by the implementation of a Levenberg-Marquardt Optimizer [23, 24] for an optimized registration. The change of the camera system to ToF cameras will also bring better results with the general ICP algorithm. So far the operation area is limited by only two cameras, because the robot’s tool center point is not detectable overall by reason of occlusions in negative y-direction. A remedy would be to place a third camera right from the robot. This is feasible with a ToF camera but challengingwith a structured light camera due to illumination disturbance from the counterpart. An alignment of 60 degrees for three structured light cameras would be better, since all the three cameras would receive the same disturbance which is less than if two of three fully receive it. A faster and more general model implementation would bring the implementation of an automatic model generation from COLLAborative Design Activity (COLLADA) [25] datawhich can be generated easily byCADprograms.With the COLLADA data (version 1.5.0) not only the geometry parameter would be loaded, the mechanical parameter as mass, inertia and center ofmass could be loaded too, which is interesting for the robot dynamic. This would remove the model preparation as mentioned in Section III for a more user-friendly application. REFERENCES [1] J. O’Neill, J. Lu, R. Dockter, and T. Kowalewski, “Practical, stretchable smart skin sensors for contact- aware robots in safe andcollaborative interactions”, in 2015 IEEE International Conference onRobotics and Automation (ICRA), IEEE, 2015, pp. 624–629. [2] C. Liu, Y. Huang, P. Liu, Y. Zhang, H. Yuan, L. Li, and Y. Ge, “A flexible tension-pressure tactile sensitive sensor array for the robot skin”, inRobotics and Biomimetics (ROBIO), 2014 IEEE International Conference on, IEEE, 2014, pp. 2691–2696. [3] A. De Luca and R. Mattone, “Sensorless robot col- lision detection and hybrid force/motion control”, in Robotics andAutomation, 2005. ICRA2005.Proceed- ings of the 2005 IEEE International Conference on, IEEE, 2005, pp. 999–1004. [4] K. Kosuge and T. Matsumoto, “Collision detection of manipulator based on adaptive control law”, in Proc. IEEE/ASME Int. Conf. on Advanced Intelligent Mechatronics, 2001, pp. 117–122. [5] C. Morato, K. N. Kaipa, B. Zhao, and S. K. Gupta, “Toward safe human robot collaboration by using multiple kinects based real-time human tracking”, Journal of Computing and Information Science in Engineering, vol. 14, no. 1, p. 011006, 2014. [6] B. Schmidt and L.Wang, “Depth camera based col- lision avoidance via active robot control”, Journal of Manufacturing Systems, vol. 33, no. 4, pp. 711–718, 2014. [7] A. Muis and K. Ohnishi, “Eye-to-hand approach on eye-in-hand configuration within real-time visual ser- voing”, IEEE/ASME Transactions on Mechatronics, vol. 10, no. 4, pp. 404–410, 2005. [8] F. Janabi-Sharifi, L. Deng, and W. J. Wilson, “Comparison of basic visual servoing methods”, IEEE/ASME Transactions on Mechatronics, vol. 16, no. 5, pp. 967–983, 2011. [9] B. Thuilot, P. Martinet, L. Cordesses, and J. Gallice, “Position based visual servoing: Keeping the object in the field of vision”, in Robotics and Automation, 2002.Proceedings. ICRA’02. IEEEInternationalCon- ference on, IEEE, vol. 2, 2002, pp. 1624–1629. [10] T.Koenig,Y.Dong,andG.N.DeSouza,“Image-based visual servoing of a real robot using a quaternion for- mulation”, inRobotics,AutomationandMechatronics, 2008 IEEEConference on, IEEE, 2008, pp. 216–221. [11] E. Marchand and F. Chaumette, “Visual servoing through mirror reflection”, in IEEE Int. Conf. on Robotics and Automation, ICRA’17, 2017. [12] D. Tsai, D. G. Dansereau, T. Peynot, and P. Corke, “Image-based visual servoing with light field cam- eras”, IEEE Robotics and Automation Letters, vol. 2, no. 2, pp. 912–919, 2017. [13] N. Shahriari, S. Fantasia, F. Flacco, and G. Oriolo, “Robotic visual servoing of moving targets”, in In- telligent Robots and Systems (IROS), 2013 IEEE/RSJ International Conference on, IEEE, 2013, pp. 77–82. [14] H. Liu, S. Huang, N. Gao, and Z. Zhang, “Binocular stereo vision system based on phase matching”, in SPIE/COS Photonics Asia, International Society for Optics and Photonics, 2016, 100230S–100230S. [15] J. Han, L. Shao, D. Xu, and J. Shotton, “Enhanced computer vision with microsoft kinect sensor: A re- view”, IEEETransactions onCybernetics, vol. 43, no. 5, pp. 1318–1334, 2013. [16] S.K.Nayar andM.Gupta, “Diffuse structured light”, in Computational Photography (ICCP), 2012 IEEE International Conference on, IEEE, 2012, pp. 1–11. [17] S. Foix, G. Alenya, and C. Torras, “Lock-in time- of-flight (ToF) cameras: A survey”, IEEE Sensors Journal, vol. 11, no. 9, pp. 1917–1926, 2011. 78