A Programming Language Extension for Probabilistic Robot Programming

Sebastian Thrun

In recent years, probabilistic techniques have led to improved solutions for many robotics problems. However, no general tools are currently available to aid the development of probabilistic robotic software. This paper presents a programming language extension to C++ that integrates probabilistic computation and learning. Its two main ideas are to make probability distributions as usable as floating-point numbers, and to smoothly integrate function approximators into C++ code. These innovations facilitate the development of robust, probabilistic robot software, as illustrated by a proto-type program for a mail delivery robot.

The full paper is available in gzipped Postscript and PDF