Seite - 25 - in Joint Austrian Computer Vision and Robotics Workshop 2020

Requirements forSearchandRescueRobots Topic Robbie Status Dimensions 112×58×120cm,ca75kg X Nr. ofOperators Onlyoneoperator required X Resistance Currently no IP certificate due toactivecooling × Autonomy Canbeeasily switchedusing GUIorcontrolpanel X Sensing Five image-streams pro- vided, as well as 2D and 3D maps X Communication Automatic drive counters connectivityproblems X CommandandControl Intuitive GUI and operator stationdeveloped X AmbientLight Can be operated during day and night due to two LED headlights X EnergyRequirements 0.36kVA X GraphicalUser Interface Three image-streams on start-up, additional 2 can be startedmanually Battery Voltage, Remaining operation time and sensor readings in status-bar X 2D map on start-up, 3D map canbevisualisedmanually X Table 2. Evaluation of Robbie’s applicability using the survey in [7] and [24]. For a description of the require- ments see table1. ing the EnRicH 2019 as a 2D, 2D with marked ra- dioactive sources and as a 3D map. As the 2D map (upper left) depicts, the calculated loop closure of the cartographer node distorted the map, resulting in sloping walls. This also led to the fact that two ra- dioactive sources were merged into one by the ra- dioactive mapping approach, since the odometry of the mobile robot is not recalculated during loop clo- sure. As can be seen in figure 4, the generated 3D Figure 4. Created maps: upper left) 2D map, upper right) mapoverlaidwithradioactivitymeasurementsbottom)3D map mapisonlypartiallydense,whichmeansthatadense 3D reconstruction was not possible in all regions. This may be because the range of the 3D depth sen- sors is too short or because the mobile robot was driven through these regions too fast. 4.ConclusionandOutlook In this paper the needs of S&R robots with re- gard to thesystemrequirementswereexaminedfrom theoperator’sPOV.Furthermore, theapproachof the UAS Technikum Vienna to implement the require- ments was examined. Search and rescue robots have to cover a wide range of application areas. Starting with the robot’s tele-operation, autonomous object recognitionandimaging,uptotheprocessingandvi- sualizationofsensordatafortheoperator. Thesearch and rescue robot of the UAS Technikum Vienna is able to generate different maps (2D and 3D), has au- tonomouscapabilities likehumanvictimrecognition or autonomous drive and has an easy to use graphi- cal user interface for the operator. The tracker base platform in combination with the robot arm and the end effector allow a high off-road mobility and offer maximumflexibility formanipulation. Future projects will deal with the tasks of tele- operation of the robot arm with the help of motion controls, the enhancement of the human recognition package by merging the already existing RGB data with point cloud data using Bayesian sensor fusion and visual servoing with reinforcement learning for optimalgripperpositioning. Further, toprovide real- world S&R capabilities it is necessary to look into possibilities towater-proof themobile robot. References [1] Arctic Robot Challenge. Arctic robot challenge. [Online]. Avail- able:https://arcticrobotchallenge.com/. [Accessed: 2019-07-17]. [2] M. Bjelonic. Yolo ros: Real-time object detection for ros, 2018. [3] J. Bohren and S. Cousins. The smach high-level ex- ecutive[rosnews]. IEEERoboticsAutomationMag- azine, 17(4):18–20,Dec2010. [4] F. Chaumette, S. Hutchinson, and P. Corke. Visual Servoing, pages 841–866. Springer International Publishing, Cham,2016. [5] B. Choi, G. Park, and Y. Lee. Practical control of a rescue robot while maneuvering on uneven ter- rain. Journal of Mechanical Science and Technol- ogy, 32(5):2021,May2018. [6] DARPA - Defense Advanced Research Projects Agency. Defense advanced research projects 25