Estimation of Pseudo-Range DGPS Corrections Using Neural Networks Trained by Evolutionary Algorithms

GPS measurements can be degraded by several type of error like unmodelled ionosphere and troposphere delays, satellite and receiver clocks inaccuracies, receiver noise, dilution of precision and multi-path effects, and also the US. military's intentionally such as Selective Availability (SA). These errors degrade the accuracy of GPS position. DGPS provides users with corrections to remove the correlated bias terms between receivers. The DGPS has the problem of slow updates. Any interruption of the DGPS service causes a loss of navigation guidance. This paper proposes the prediction of DGPS corrections using Neural Networks (NNs) trained by evolutionary algorithms such as the genetic algorithm and the Particle Swarm Optimization (PSO). A low cost commercial and single-frequency GPS receiver is utilized to demonstrate the DGPS positioning. The experimental results show the feasibility and effectiveness of the proposed methods. The results are analyzed and compared with NNs trained by the back propagation algorithm. The NNs trained by PSO gives better accuracy in estimating the DGPS corrections; so that the total RMS error reduces to less than 1.32 metre with SA on and 0.46 metre with SA off. Copyright (C) 2010 Praise Worthy Prize S.r.l. - All rights reserved.