Distributed Object Recognition

Would it be not wonderful if we can ask a robot (a personal assistant) to locate an object that we have misplaced and the robot can reply back without moving from its place?

The above question is the ultimate aim of this project. I worked on The Corpora's Q.bo, a fairly advanced robot with great human interaction capabilities (Capable of recognizing voice commands and can detect faces).

This project is a part of the larger project by the various faculty of Örebro Universitet at Ängen (AASS Research facility), Örebro, Sweden. The aim of the Ängen project is to create a better and easier way of independent living for the elderly people.

My work primarily focuses on the way we interact with the robot to ensure that the robot (Q.bo) can locate a remote object. Q.bo was also interfaced with the central workstation so that it's learning can be used by various other smart devices or robots in the apartment. When asked to locate an object, Q.bo accesses the various cameras in the apartment and tries to locate the object in question within those camera images. The results are then conveyed to the user via voice. If the robot experiences ambiguity i.e if more than one instance of the object is found, then the robot will convey the same to the user and the robot will also be able to discuss about the features of the objects found and a conversation can be held with which it can resolve the ambiguity and the user will be able to find out the location of the object he is looking for.