Robots can assist us in many scenarios that are inconvenient or unsafe for us, e.g., when social distancing needs to be maintained, or under the extreme conditions imposed by hazardous industrial scenarios and space explorations. Before we can safely delegate tasks to robots, however, we need to ensure that they can accurately make sense of their environment. That is, we expect robots to be capable of correctly interpreting the high-volume and diverse data that are coming through their sensors. Specifically, in my doctoral research, I tackle the problem of robot sensemaking from the standpoint of Vision. Therefore, in this session, I will introduce this challenge and present the work I have done in this area, in particular by showing how the integration of a Deep Learning approach to object recognition with common-sense knowledge helps to improve significantly the sensemaking capabilities of Visually Intelligent Robots.

Watch the webcast replay >>