Realistic humanoid robots have emerged in the last two decades but the emotional intelligence of these machines has been limited. To teach humanoids how to emotionally communicate with humans, researchers have been increasingly relying on machine learning algorithms. While the software used to implement machine learning algorithms is largely open source, facially expressive humanoid robots are expensive and inaccessible to most people, thus limiting the number of researchers in this field. This paper aims to aid potential artificial intelligence researchers by providing a relatively inexpensive, open-source robot that can serve as a platform for research into emotional communication between humans and machines. Eva, the robot described in this paper, is an adult-sized humanoid head that can emulate human facial expressions, head movements, and speech through the use of 25 muscles, including 12 facial muscles that can produce a maximum skin displacement of 15mm.

This work has been supported in part by US National Science Foundation (NSF) grant NRI 1925157