The application of learning algorithms in the development of natural interaction

This paper argues that natural interaction with a machine can be realized and improved by using learning algorithms. Through the use of supervised and reinforcement learning algorithms, a robot was created that can be trained to perform actions using only verbal commands. The user has complete freedom in choosing preferred commands and what actions should be linked to these commands. The combination of supervised and reinforcement learning resulted in a fundamentally different way of interaction with a robot. The way this system was set up can be used as a framework for new projects, giving designers a new tool to improve human-machine interaction.