Low Cost Human - Machine Interaction Platform for Bridging the Uncanny Valley, Using Humans as the Control System Input.

Abstract

This master's thesis describes the development of a conceptual prototype of a smooth actuation and control system of the eyes in a healthcare simulation mannequin with the aim to achieve perceived human like eye movement and behavior. The chapters of this thesis identifies, collects, judges and conceptualizes technologies to simulate the eye movement and behavior. The project identifies and develops a system that achieves realistic human-machine interactions, without a feeling of eeriness or unease, as is common with humanoid simulations.

Prototypes are shown that tackle the problems of human-like movement and behavior. The system prototype allows the operator to look through the eyes of the mannequin by displaying the mannequin's view to the operator in the head mounted device. The head mounted device processes and captures the eye movement of the operator and mimics this in mechanical actuation of the eyes of the mannequin. By letting the operator see the environment around the mannequin, the operator is able to interact with medical personnel through the mannequin.

Possibilities for automation of the behavior has been proposed, where the most promising is supervised learning in artificial neural networks. By using deep learning, it may be possible to train an algorithm to mimic human behavior and thus removing the need for operator control.

The system that has been proposed is an alpha prototype that has implemented critical functions for operation and proof of concept. Implementation is dependent on further work and optimization. The system has the potential to provide further dimensions of simulations by introducing nonverbal communication and diagnostic tools that use the eyes.