Autonomous Mobile Robot with Associative Hebbian Learning (1993)
As a final year project for my engineering degree, this robot was a lot of fun to build. I still have the old report that I can post for completeness, but in summary this was a robot that you could really hurt - ultrasonically speaking. The neural network circuitry would happily drive the vehicle towards the light guided by the two sensors mounted at each end, but zap it with an ultrasonic beam and it would run for its life. It even had enough brain cells (7 to be exact) to be able to associate the ‘pain’ zap with the light. Sadly this dog was born to live on the wrong side of pavlovian learning curve.
Ph.D. Thesis : Intelligent Autonomous Mobile Robot Navigation (1996)
So if it takes seven neurons to create pain, how many does it take to run away from it? To answer that you need to navigate, and what better creature to ask than a rat. It turns out that rats use ‘place cells’ buried deep in the hippocampus that fire repeatedly in familiar locations which they then use to track their locomotion through their environment. Through a combination of simulation and building some real robot hardware it was possible investigate learning ‘on the fly’ based on biologically inspired models of learning.
Robot Explores the Maze Neural Map Learned Through Experience
The Robot Plans its Path Fast Navigation to Goal
Yale University Post-Doctoral Research Robot Helicopter - 3D Pictures (1999)
In 1997 I was invited to Yale University’s Center for Systems Science to continue my robotics research as a post doctoral researcher. As part of that work I developed an autonomously controlled R/C helicopter by building out a flight ready control system built from a suite of sensors, actuators, and all the required computational infrastructure.
To view the helicopter in glorious 3D, position yourself infront of the following two images about 10 inches away from the screen. Then cross your eyes so that both images overlap whilst keeping them in focus. If you can hold that for more than a moment your brain will merge the slightly different images together to see the helicopter in full 3D. Enjoy!
