The Systems Journal Club looks into some of the details of self-driving vehicles. Danny Smith will moderate the discussion of several projects.
First, the pioneering work of Williams alum, Dean Pomerleau ’87, who helped to build very early self-driving vehicles at what is now CMU’s Robotics Institute. We’ll look at two of Pomerleau’s papers:
ALVINN, an autonomous land vehicle in a neural network, 1989 CMU Technical Report
and
RALPH: Rapidly Adapting Lateral Position Handler, 1995 Symposium on Intelligent Vehicles
Danny may also discuss the architecture of Google’s self-driving car, a project run by Sebastian Thrun, and others. You can read more Google’s efforts at the IEEE Spectrum site: http://spectrum.ieee.org/
Projects like these are examples where AI approaches to problem solving are tested in real-world applications. Increasingly, systems we build to will need to depend on encapsulating algorithms that
produce approximate results. We need only look as far as automated mapping, translation services, voice-recognition systems and Siri to understand some of the typical challenges.
In 1995 Pomerleau’s RALPH project drove a vehicle from Washington DC to San Diego (2800 miles). This was the first long-duration test of a self-driving car (Google was founded in 1998). It cost $20,000 to build.