David Bullock - mail@davidbullock.uk

II. Robots - Technical Overview

I chose to build my robots using the Lego Mindstorms Robotic Invention System as this seemed to offer the greatest flexibility. Mindstorms was originally developed by Massachusetts Institute of Technology (MIT) in conjunction with Lego as an education tool to help teach programming to children. Because it is based upon Lego building blocks it is an ideal tool for robotic experimentation. New ideas and prototypes can be built and modified quickly and easily. Consequently Lego Mindstorms systems can be found in many leading robotic research establishments.

My own experiments were based around autonomous movement and fall into two basic categories: 1. free-roaming and 2. light reactive, maze-solving. The first of these focused on creating a robot that could move freely around an enclosed space and circumvent any obstacles encountered. I used this robot as the basis for the Cambot, a robot fitted with a digital camera which could film its own progress. I later modified the programming to get two roaming robots to dance synchronously. The second challenge was to create a robot that was able to negotiate a maze made out of insulating tape stuck on the floor. To achieve this the robot not only had to detect the maze edges using light sensors but also have a strategy to recognise and deal with dead-ends and other obstacles.

Cambot

I wanted to create a free-roaming robot that could film its movements primarily as a means to generate digital video footage autonomously without any human intervention. This is an extension of a number of ideas which I have wanted to explore further prior to this project. I have long been fascinated with the concept of self-generating artwork – a kind of autopoiesis where the creative act is either performed mechanically or inadvertently by the intervention of a third party such as a viewer. A robot that could hold a camera while roaming randomly got over an essential problem which was how to create a random film.

With the dancing variant, one robot performed one of six movements randomly and the second robot copied this movement. This was achieved by the first robot sending a random number to the second robot. Upon sending/receiving a number each robot would then execute the movement corresponding to the number sent. It should be noted that each robot did the exact opposite of the other robot, thus when one moved left the other moved right, when one moved forward the other moved backwards. As each movement lasted the same length of time a sequence of movements strung together would be rhythmical with each movement defining a beat. Again my interest in this stemmed from the idea of creating an automated, random dance.

See Cambot test videos
see Cambot building instructions*

Mazebot
Following my trip to Crete (home to the most famous mythical labyrinth) during the summer, I realised that there was a connection between my robot experiments and the videos I had been working on. Robots are commonly built to solve mazes as a demonstration of robotic AI and upon reviewing my earlier footage it struck me that some of it was reminiscent of being in a labyrinth or maze. After making this connection I intended to use the Cretan legend as a motif to hang my work on robots and video together. As it happened I ended up dropping the Minoan labyrinth theme as I came to realise that I was always more interested in exploring ideas of free will, determinism and metaphoric boundaries. Maze-solving robots and mazes are just as relevant in these contexts.

Technically this was an interesting problem to solve. I came to understand that there are no absolute solutions and that success was relative. Design success or failure could only be measured by comparison to previous attempts. A better design increases the likelihood of success and a worse design decreases the likelihood. Building in this way on successful strategies and rejecting less successful strategies the design process effectively became evolutionary. I also think it was analogous to learning new skills within intelligent species such as humans. When learning a new skill failure is likely to be high to begin with. By observing which strategies bring failure and which success behaviour can be modified accordingly. Highly skilled individuals are therefore those that have been through this iterative process more times or are able to analyse and adapt more easily.

See Mazebot test videos
see Mazebot building instructions*

* = opens in new window