Page - 163 - in Proceedings - OAGM & ARW Joint Workshop 2016 on "Computer Vision and Robotics“

chairs and small items did not usually have a significant influence on the localization quality, as long as discrepancies with respect to the static map were limited. Removing or changing furniture that provided relevant geometrical references present in the map, however, led to serious differences, es- pecially in narrow areas and corners with little space to recover while moving. Detecting when the robot isactually lostandtryingtoapplyautonomousrecoverymethodsareverychallengingproblems that require specific research. 6. Functionsusage Table 1 provides results on how often different functions were used in the user trials conducted in Vienna. Preliminary results show that most of the users evaluated these functions overall as “Good”, but there were also important failures and negative comments. Results in the lab were significantly betterandwethinkthat likelyreasonsarerelatedtothefollowingfacts: therewasmoretimeforsetup, preparationandimprovements; theenvironmentiseasierandbetterknown;weknowthesystembetter than the real users; robot transportation and long term operation can degrade platform performance andcalibrations, etc. Table1. FUNCTIONS USAGEDURING THETRIALS User GoTo CallRobot Recharge Used Cancelled Used Cancelled Used V1 89 33 340 190 49 V2 29 8 172 81 186 V3 18 5 74 21 126 V4 46 26 97 75 16 V5 117 24 71 42 88 V6 386 85 146 60 312 V7 41 17 349 263 168 7. Conclusionsandfuturework This paper presented a summary of contributions and findings for navigation tasks of care robots operating in real home environments, using an RGB-D based sensor setup. Increased flexibility and adaptability over existing solutions were provided for the tasks execution, also addressing different limitationsof the sensorsused. Our experience indicates that autonomous RGB-D navigation in real home environments is usually feasible and is much appreciated but presents drawbacks, open problems and further challenges. Di- rections for futureworkwerealreadyhighlighted in thepaper. References [1] R. Ambrus, N. Bore, J. Folkesson, and P. Jensfelt. Meta-rooms: Building and maintaining long term spatial models in a dynamic world. In Proc. of IEEE/RSJ International Conference on IntelligentRobots andSystems (IROS), 2014. 163