Development of Humanoid Robot Head Based on FACS

Abstract

Advances in artificial intelligence have allowed humanoid robots to communicate and serve human needs. However, displays are often used for humanoid robots to perform expressions, limiting the authenticity and naturalness of human-robot interactions. This paper presents the design and analysis method of the humanoid robot head to improve interoperability. At first, the Facial Action Coding System (FACS) was used to determine which Action units were required to perform expressions. Control points are required to generate Action units, and they are determined based on the action of facial muscles. In the next step, the humanoid robot head is divided into 3 parts: eyes, jaw-lip, and neck. Each section was dug into anatomy to research the shape, size, and range of motion of bones and joints. Mechanisms are designed based on this data to realistically simulate the structure and motion of individual parts of the human head. In addition, the forward and inverse kinematics problems of each mechanism are also solved to determine the control equation of each control point and workspace. These results demonstrated the theoretical operability of mechanisms that are suitable for simulating human facial motions. Finally, a humanoid robot head is implemented based on this method to evaluate the capability to simulate motion and perform expression. Experiment results show the simulation rate is up to 94.72%, and the expression recognition rate is 89.93%