Bio-inspired robot perception coupled with robot-modeled human perception

Research statement – part 1: My overarching research goal is to provide robots with perceptional abilities that allow interactions with humans in a human-like manner. To develop these perceptional abilities, I believe that it is useful to study the principles of the human visual system. I use these principles to develop new computer vision algorithms and validate their effectiveness in intelligent robotic systems. I am enthusiastic about this approach as it offers the dual benefit of uncovering principles inherent in the human visual system, as well as applying these principles to its artificial counterpart. Fig. 1 contains a depiction of my research. Perspective-taking: In our everyday lives, we often interact with other people. Although each interaction is different and hard to predict in advance, they are usually fluid and efficient. This is because humans take many aspects into account when interacting with each other: the relationship between the interlocutors, their familiarity with the topic, and the time and location of the interaction, among many others. More specifically, humans are remarkably good at rapidly forming models of others and adapting their actions accordingly. To form these models, humans exploit the ability to take on someone else’s point of view – they take their perspective [6]. In [7], we introduced an artificial visual system that equips an iCub humanoid robot with the ability to perform perspectivetaking in unknown environments using a depth camera mounted above the robot, i.e. without using a motion capture system or fiducial markers. Grounded in psychological studies [19, 12], perspective-taking is separated into two processes: level 1 perspective-taking comprises the ability to identify objects which are occluded from one perspective but not the other; this was implemented using line-of-sight tracing. Level 2 perspective-taking refers to understanding how the object is perceived from the other perspective (rather than just understanding what is visible from that perspective; see [19]); this was implemented using a mental rotation process. While [7] implements perspective-taking in a robotic system, it does not provide any insights into the underlying mechanisms used by humans. In [8], we investigated possible implementations of perspective-taking in the human visual system using a computational model applied to a simulated robot. The model proposes that a mental rotation of the self, also termed embodied transformation, accounts for this ability. The computational model reproduces the reaction times of human subjects in several experiments and explains gender differences that were observed in human subjects. In future works, I would like to explore the relationship between perspective-taking and active vision. Active vision 0 20 40 60 80 100