Model-Based Environmental Visual Perception for Humanoid Robots

The visual perception of a robot should answer two fundamental questions: What? and Where? In order to properly and efficiently reply to these questions, it is essential to establish a bidirectional coupling between the external stimuli and the internal representations. This coupling links the physical world with the inner abstraction models by sensor transformation, recognition, matching and optimization algorithms. The objective of this PhD is to establish this sensor-model coupling.