Department of Computer Science and Electrical Engineering

Modelling of an Autonomous Doppler Navigation System Sensor Input Processing Unit

Paul V. Petraitis

Abstract:
Modern carn navigation systems use signals from GPS satellites to determine their position.  If contact with the satellite signal is lost, an autonomous system that does not rely on external signals must assume navigation responsibilities.  Current autonomous systems use expensive gyroscopes.  A doppler radar implementation, with sensors pointing at the road at each corner of the car, would be much cheaper.  However, signals received by the doppler sensor experience signal spreading, asymmetry and ray scattering by the road surface.  This makes it difficult to determine the exact doppler frequency, and hence speed, of the vehicle.  This model, presented in Matlab, proposes a novel method for extracting speed information using cross correlation of the detected power spectral density with an ideal distribution.  With the speed of each sensor known, the veolcity of the vehicle can be determined by geometry.  Limiting analysis to circular trajectories, the model obtains low average errors in position but at relatively high computation times.  The computation time can be reduced but at the cost of increased error.  The implementation of this model on a dedicated DSP would reduce computation time and is proposed for future investigation.

Complete thesis:
thesis.pdf
conference_paper.pdf

Additional material:
Matlab code

About the Author


Dept of Computer Science and Electrical Engineering / Paul Petraitis / pvpx@yahoo.com    last mod 03/09/98