A mini-robot using self-learning for navigation

Abstract

The robot described in this thesis is aimed at being used for research on machine learning. The idea was to provide an interface to the physical world to a neural network running on a host computer. This thesis describes the background literature, designed robot, as well as a charging station which enabled the robot to recharge itself. The main focus of this work was to build a robot as small as possible, without designing custom actuators or batteries. The final size of the robot is of 2.7 cm3, which is much smaller than the smallest robot designed so far with significant sensing capabilities(TinyTERP with extension board, 7.65 cm3), to the best of my knowledge.