Abstract
This paper presents a system enabling a mobile robot to autonomously pick-up objects a human is pointing at from the floor. The system does not require object models and is designed to grasp unknown objects. The robot decides by itself if an object is suitable for grasping by considering measures of size, position and the environment suitability. The implementation is built on the second prototype of the home care robot Hobbit, thereby verifying that complex robotic manipulation tasks can be performed with economical hardware. The presented system was already tested in real apartments with elderly people. We highlight this by discussing the additional complexity for complete autonomous behavior in apartments compared with tests in labs.
Highlights
Robots have been envisioned as helpers at home for a really long time
Great advancements are achieved every year but, if we look closer, a general multi-purpose autonomous robotic butler is still far away: it is an ambitious goal [1]–[5]
We address the whole process: starting the pick up command via pointing gestures, integrating autonomous navigation to the selected area, incorporating methods for the robot to find the object and including a checking stage to assess whether the object size, properties and position are adequate for a grasping task
Summary
Robots have been envisioned as helpers at home for a really long time. In this direction, great advancements are achieved every year but, if we look closer, a general multi-purpose autonomous robotic butler is still far away: it is an ambitious goal [1]–[5]. The first prototype version of the Hobbit robot for multiple tasks and results from laboratory trials were described in [13] This version included a preliminary set of functions to check whether an object should be grasped or not but the limited arm kinematics and a slow method to decide if the task was successful or not, together with limited navigation capabilities, resulted in a not very impressive overall performance. This work included a high level general overview of all components and functionalities of the system It presented the lessons learned from field trials conducted with real users in their own homes for up to three weeks, mostly from a HumanRobot-Interaction perspective. Thereby we focus on trial runs in which the robot should decide not to grasp, since safety for the user and the robot itself is a main issue for robots autonomously operating in real human environments
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
