IDSIA Meeting Room, Tuesday at 14:30

Speaker: IDSIA's new postdoc Ricardo Omar Chavez Garcia

Title: Passive and interactive perception in robotics

Abstract: In this talk I present two perspectives of the perception task in robotics. First, a passive approach to detect, track and classify multiple objects in driving environments. Such approach relies on multiple sensor fusion to provide evidence for object classification and considers uncertain observations for state estimation. Second, a probabilistic approach for learning relationships between an agent’s motor primitives, its perceptual capabilities, and the changes generated by its interaction with the environment. This approach represents the correlation between action and its resulting changes in the perceptual state of the scenario and follows a developmental methodology.