Page - 54 - in Joint Austrian Computer Vision and Robotics Workshop 2020

for these systems are identified and requirements are defined to eliminate or sufficiently reduce the risks associated with thesehazards. Thestructuredprocess fromtheMachineryDirec- tive down to the EN ISO 10218-2 shows that every safety-relevant change of the application requires a renewedriskassessment,unlessthishasalreadybeen considered in the original assessment. New risk as- sessments on an already existing work system might make required modifications impossible due to lim- itedflexibilityinthedesignorre-design. However, in order to consider safety-related changes in the orig- inal assessment, prospective modifications and thus systemvariantshave tobeconsidered inanearlyde- sign phase. For this approach, however, the link be- tweenmodificationsandsafety-relatedaspects isnot yet clearly explored. Even though, the Technical Specification for col- laborative robotic applications, ISO/TS 15066:2016, presents a correlation of the applied robot’s safety mode and the system’s respective safety-related changes, thesafetymode isonlyoneofmanysafety- relevant modification dimensions within human- robot work systems [2]. Further, it shows draw- backs in applying the proposed safety measures, es- pecially when integrating heavy industrial robots or sharp objects or estimating the human approach ve- locity [3]. Additionally, there is no advice consider- ing the robot’s movement predictability due to col- lision avoiding path planning or varying task alloca- tion patterns. Assafetymodesmightnotbeanappropriateclas- sification scheme for the identification of safety- relevant changes, new classifications schemes have been introduced, such as in [4, 5]. However, af- ter an extensive literature review, [6] came to the conclusion that classification schemes for collabora- tive human-robot work systems are not applied con- sistently, which may lead to an incorrect identifica- tion of safety-relevant changes. To counteract this, model-based approaches have been developed either basedonformalmathematicalmodels, suchas in[7], orbasedonsimulationmodels, suchas in [8]. Arisk management simulator was for example introduced in [9], whereas [10] introduced a task-based char- acterization of human-cobot safety. Further, a met- ric depending on the distance between robot and hu- man as well as the robot’s structure was introduced in [11]. However, none of the proposed approaches con- Area A Area B conveyor belt safety mat UR3 UR10 m ob il e m an ip ul at or feeder Figure1.Structureof the workplace. siders a mutual influence of modifications on safety. In this sense, a structured procedure with the aid of an Morphological Box (MB) was developed and is described in detail in the following paper. The pro- posed approach should support manufacturers and system integrators in the consideration of safety- relevant changes in an early design phase of the planned collaborative human-robot work system in- cluding itsprospectivemodifications. 2. ImpactofModifications Within the DR.KORS project on dynamic recon- figurability of collaborative robotic systems, 50 di- mensions of modifications were identified directly influencing the safety of a human-robot system. The modification dimensions can be classified in work- piece, end effector, contact points (between human and robot), speed / acceleration, task / workflow and operatingconditions/changeofplace. Theimpactof modificationdimensionswillbepresentedona labo- ratoryusecaseexampleforassemblingrocker levers. 2.1.UseCaseDescription In the laboratory use case rocker levers consist- ingof threeseparatecomponentsareassembledwith a collaborative human-robot work system. Adjust- ing bolts are mounted on two separate rocker levers which are then assembled on a trestle. At this point, the positioning of the rocker levers on the trestle needs manual dexterity as the components tilt eas- ily. Rocker levers and trestles are provided either by feeders or on a conveyor belt. The manipulation, 54