The first session after the invited speaker was on Robotics. All papers in this session related to humanoid robots, and the Nao in particular. Mike Ligthart presented his work on trusting robots in a very refreshing way. He argued that it is very important to be aware of the extend that humans trust robots that they work with, and then reported on his experiments where people were asked to report the degree of trust they had in a robot that was trembling. He found intriguing differences in these reports between an experiment with a video and an immersive (virtual) environment that he could not (yet) account for. Further study perhaps is not just leading to insights on trusting robots, but also on how the process of trusting works for people in general.
The second talk was by Zhenglong Sun who explained how the process of bipedal walking can be modeled using a pendular model, and he reported on how this affected the energy efficiency of a Nao robot. After his talk there were questions on how the speed and efficiency compare to other existing models than the default Nao implementation; an interesting direction for future work!
The session was rounded up by an inspiring talk by Chiel Kooijman on optimizing walking parameters for the Nao in robot soccer. This multi-objective optimization problem was approached using evolutionary programming. Given the promising first results on a simulation of the Nao, there are many issues left, and thus many directions for follow-up work. For example, how to let the robot safely learn the parameters in the real world? Which parameters to learn? How to deal with the multiple objectives? How to make the parameters context/task dependent? This talk again shows how the Robocup is an inspiring domain for AI research.
– Mathijs de Weerdt
