Modeling and Information Processing
TAP - Advanced Particulate Techniques
GPS reception
Particle filtering (PF) is a non-linear filtering technique based on Monte Carlo approximation. Its aim is to estimate the state's conditional density recursively (phase, phase drift, pseudo-distance) from the measurements (GPS signal) without performing linearization. Theoretically, PF allows for phase tracking and therefore the execution of a long and coherent integration of the GPS signal.
The Global Positioning System is used to locate a receiver accurately on earth via a constellation of satellites. Each satellite generates two frequency-modulated carriers per navigation message and a PRN (PseudoRandom Noise) code, for example the C/A (Coarse/Acquisition) code. The signal received by the receiver is weak and is drowned by the ambient noise. The receiver generates this PRN itself to correlate it with the signal received. The correlation peak gives an estimate of the delay and therefore the distance (pseudo-distance). Starting from three satellites that give three pseudo-distances, the receiver can estimate its position (it needs four satellites if it wishes to correct the drift of its own clock). However, the signal integration algorithms assume the delay is steady during the duration of the PRN code. For example, the Doppler frequency is not estimated. In the case of interference with the GPS signal, we must be able to integrate longer and therefore model and estimate the signal more finely. The contribution of the PF techniques developed at Onera/DTIM for GPS has been the subject of a study validated by a full scale evaluation.
Convergence of positioning errors by the particle filtering algorithm.
