Spatio-temporal representation for cognitive control in long-term scenarios

The FP-7 Integrated Project STRANDS [1] is aimed at producing intelligent mobile robots that are able to operate robustly for months in dynamic human environments. To achieve long-term autonomy, the robots would need to understand the environment and how it changes over time. For that, we will have to develop novel approaches to extract 3D shapes, objects, people, and models of activity from sensor data gathered during months of autonomous operation. So far, the environment models used in mobile robotics have been tailored to capture static scenes and environment variations are largely treated as noise. Therefore, utilization of the static models in ever-changing, real world environments is difficult. We propose to represent the environment’s spatio-temporal dynamics by its frequency spectrum.