Being a Roboticist naturally means that I spend an awful lot of time fiddling with robots, so on this page I've collected a brief description and some photos of all the robots I've worked with so far, in reverse chronological order (newest first).

I should probably make this clear: I'm a scientist, not an engineer, so although I'm a Roboticist I mostly don't build robots. Building robots is fun, cool, and hard (hats off to people who are good at it), but what I do, mostly, is make robots do interesting things.

As part of my work on the Hydra project, I developed portions of the propulsion control microcontroller software on the Hydron units, initially working with the first prototype (pictured on the right), and then later with the batch produced robots (below, picture courtesy of Tim Taylor).

A Hydron unit is a roughly circular robot intended for use in collective tasks under water. It is is actuated in the horizontal plane by four nozzles which expel water drawn through an impellor at the bottom of the unit, with a rotating collar selecting the active nozzle. A syringe draws or expels water through the bottom of the unit to control unit buoyancy, actuating the unit along the vertical axis. The remarkable mechanical design of the Hydron was developed by Ali Yener Boztas.

Each unit's hull also supports a small set of switchable optical sensors and emitters for data transmission, developed by Lukas Lichtensteiger.

I used Dangerous Beans for my MSc project at Edinburgh, where I used it to implement the reinforcement learning model I developed. The robot had to navigate an artificial arena, learn a distributed topological map of the arena, and then perform reinforcement learning over the map. Reinforcement was based on three internal drives that motivated the robot to find a food puck, explore the arena, and return home. Details can be found in my MSc on my publications page.

Dangerous Beans was a standard K-Team khepera with a pixel array turret, and stood about 7cm high. The robot had eight infra-red sensors (used for obstacle avoidance and wall following), two incremental wheel encoders (used for dead-reckoning position estimation) and a 64-pixel linear vision array (used for puck spotting).

The arenas built for Dangerous Beans measured roughly 90cm squared, and were built on a solid piece of wood with styrofoam, cardboard, drawing pins and insulation tape.

Dangerous Beans was named after a character in Terry Pratchett's brilliant The Amazing Maurice and His Educated Rodents. I developed a similar control system (this time simulated under Webots) for an extension of my MSc work, and called the new robot Peaches 'n Cream.

Ringo was the first of two robots that Logi Pettursen and I designed as part of the Intelligent Sensing and Control course at Edinburgh.

Ringo was built using Edinburgh's brain brick kit and lego technic parts, and was simply meant to avoid obstacles and not get stuck while wandering around the ISC lab.

Ringo had four bump sensors (built into its sensory ring) and two whiskers for obstacle avoidance, and a Hall Effect sensor for determining when it was stuck.

Ringo had quite a neat design - its two orthogonal axial pieces were joined in the middle, the brain brick was placed on top of that, and then its sensory ring or "shell" was mounted on top, anchored at the end of each axial piece.

We decided on the name Ringo because the fully assembled robot looked like a beetle.

Os was the second ISC robot that Logi and I built at Edinburgh. Like Ringo, Os was built out of Edinburgh's brain brick kits, but it was required to play "robot rugby" rather than simply avoiding obstacles.

Os had two wheels with independent motors and a jaw with a motor for trapping balls. It also had two forward-facing bump sensors used for obstacle avoidance, a whisker sensor for detecting when a ball was between its jaws, and a pair of wires underneath it for detecting silver tape. We included two infra-red sensors at different heights that were intended use for spotting a ball, but we never quite managed to get it right.

Os was placed in an arena that was divided into three parts, and was required to find softballs in its arena and move them out of it. The arenas were bounded either by solid walls or by a strip of silver tape on the floor.

Unfortunately in the end Logi and I didn't have enough time to get Os quite up to scratch; it mostly worked, but not well enough to get to the finals of the robot rugby competition.

Even though I know virtually nothing about rugby, Os was named after an (apparently quite good) South African rugby player.

Hurtle was the first robot I ever programmed. It was built from a Lego Mindstorms Kit and Mindstorms Vision Kit by Dylan Shell when we were at Wits together. The kit belonged to George Christelis and I, so we brought it to Edinburgh with us and reconstructed it.

Hurtle is a two-wheel drive buggy with forward steering and a camera mount that can rotate horizontally, and runs brickOS (formerly legOS). It is an extremely well designed robot - it is the most functional, robust and reliable piece of lego construction I have ever seen. The intention was to make the camera and serial control wireless (the Mindstorms infra-red tower is short ranged and line of sight) and use it as a remote controlled robot. I've written the control software for the RCX and a C library for this and obtained a wireless camera, but I still need to get a capture card and build an RF serial port.

Hurtle's lack of sensors (only the video is really useful) make it difficult to use as a research robot. Edinburgh's brain brick kits are far superior to the Mindstorm kits.