In 1989, ALVINN, the Autonomous Land Vehicle In a Neural Network, drove itself around Carnegie Mellon campus for the first time.
By the early 1990s, it was able to hit 70mph, relying on a CPU the size of a fridge.
A video of the event recently surfaced on Twitter in a conversation between Dean Pomerlau, the CMU professor behind ALVINN, and Oliver Cameron, who heads up a self-driving project at Udacity.
Cameron told The Verge that ALVINN could be seen as the forefather of today’s self-driving cars:
“Why? The approach ALVINN took was using a neural network to drive the car, which was absolutely novel for the time and is quickly becoming an increasingly popular approach with self-driving car efforts.”
A car relying on neural networks can set off from any location, even if it’s never seen it before, Cameron added.
Pomerlau’s project was the result of eight years of research at CMU’s robotic institute funded by the US military.
The “neural network powering ALVINN was beautifully implemented, but constrained very much so by the hardware”, Cameron wrote on Medium.
Pomerleau is now offering Cameron’s students a few tips about lessons he learned from the ALVINN project, The Verge reported.