Page - 130 - in Proceedings - OAGM & ARW Joint Workshop 2016 on "Computer Vision and Robotics“

human state evaluation, and safety standards. These challenges are addressed in the 4-year research project ”CollaborativeRobotics” (CollRob), launched in20151. In this article, anoverviewabout the backgrounds,objectivesandcurrent statusof thisproject isgiven,whichshall serveasareferencefor proceeding publications and research initiatives. Chapter 2. describes the elaborated hardware and software architecture of the overall CollRob system. Chapter 3. presents different specified levels of complexity for human-robot collaboration as well as envisioned use cases to test developed concepts and methods. Chapter 4. outlines concrete research challenges addressed within CollRob. Finally, Chapter5.givesaconclusion. 2. RobotSystemHardwareandSoftwareArchitecture To enable collaboration between a human and a robot, a variety of sensors is required. Figure 1(a) shows an overview of the CollRob hardware setup. Besides the robot itself, which is equipped with torquesensors,weusededicatedvisualandproximitysensors for i)monitoringof theworkspaceand ii) detection and tracking of the human. In addition, we use wearable bio-sensors and eye tracking glasses to monitor the behavior and state of the human. A tablet, augmented reality (AR) glasses, microphones, speakers, andgesturebracelets areused ashuman-robot interactiondevices. The broad range of different sensors and devices results in complex data-flows, which in turn cause strong dependencies and couplings between the individual application parts. A software architecture for an application like this needs to relax the strong dependencies in order to enable re-usability, scalabilityandeasyexchangeof individualmodules. Publish/subscribehasproven tobeawell suited architectural pattern to accomplish this decoupling. An application is composed of modules which provide data (publish) and consume data (subscribe) from others. However, the transport of this data ishandledbyadedicatedinfrastructuresuchthat themodulesneednotknowprovidersandconsumers of their data. Figure 2 shows the modules which compose the CollRob system. In our architecture, each module is modelled as publisher and/or subscriber. We realize the publish/subscribe system using theRobot OperatingSystem (ROS). 3. LevelsofComplexityofHuman-RobotCollaborationandUseCases Within the CollRob project, different levels of complexity of collaboration have been specified (see Table 1). The category A is in fact non-collaborative. However, taking it into account can be useful for setting up the general CollRob system architecture before going into collaboration details. The categories B to D consider human-robot interaction with gradually increasing complexity. Category E describes the case of two collaborating robots (or one robot with two arms) and category F the interactionbetween tworobots andoneormorehumans. category A B C D E F umbrella term encapsulation H-R co-existence static H-R collabora- tion dynamic H-R collabo- ration static/ dynamic R-R collaboration static/ dynamic H-R-R collaboration interaction level interaction-free opera- tion safety stop static collaboration dynamic collaboration static/ dynamic R-R collaboration static/ dynamic H-R-R collaboration actors robot human+ robot human +robot human +robot 2 robots 2 robots+human(s) temporaldependence independent interrupt sequential simultaneous sequential/simultaneous sequential/simultaneous spatialdependence separated separated shared shared shared shared human-robotcontact none rudimentary pronounced comprehensive n.a. pronounced/comprehensive Table1.Overviewof interactioncategories Concerning the described categories, different application domains will be addressed in CollRob 1http://www.joanneum.at/en/robotics/reference-projects/collrob.html 130