Abstract

This is a report of a semester long project I did at the University of Auckland between March and July, 2005. You can view the report by viewing the pages in this category or by downloading the PDF below.

Report Abstract

As the number of robots in the world increases, from automatic vacuum cleaners, to toy robot dogs, to autonomous vehicles for the military, the need for effective algorithms to control these agents is becoming increasingly more important. Conventional path finding techniques have relied on having a representation of the world that could be analysed mathematically to find the best path. However, when an agent is placed into the real world in a place it has not seen before, conventional techniques fail and a fundamentally different approach to path finding is required. The agent must rely on its senses, such as the input from a mounted camera, using this information to get around. In this project, a virtual city is implemented, and agents must navigate their way around it by using only what they see from their point of view. By feeding what the agents see into a neural network, they are able to learn how to avoid obstacles, follow the road, and they also show promise in using this technique for path finding.