This project was implementing Markov Chain Monte-Carlo Localization (MCL) for a field robot. MCL is a particle-filter approach. At that time (2001), MCL had only been used on indoor robots with low-resolution sensors (such as ultrasound.) This paper describes my highly efficient implementation based on road detection, intended to be deployed on the General Dynamics XUV. I made some changes to the published algorithms to improve performance in this application, and also added a new (and highly effective) procedure to validate any apparent position fix based on reprocessing recent measurement according to the position hypothesis. I later did successful tests on NavLab11 using an actual road sensor, and also using elevation map data. Elevation can be sensed using a barometric altimeter, though I used GPS for testing.