Software Development:
Massive, Rapid Network Processing
with Ambiguity Resolution
Geoff Blewitt
Problem


Desirable to include as many stations as
possible to define SNARF
Desirable to produce one unique, rigorous
solution with ambiguity resolution

PPP is fast, linear ~N (<10 sec/station/cpu)

But ambiguity resolution is slow (~N4)


Typically limited to ~50 station clusters

Cluster processing not rigorous, not convenient
Can ambiguity be made linear and yet rigorous?
Solution

Ambizap algorithm – input is PPP solution

Solves ambiguities for N-1 neighboring pairs

So is linear ~N, (< 5 sec/station/cpu)

Does not count data twice

Agrees with full network ambiguity resolution

To << 1 mm rms in station coordinates

Implemented for cluster processing

700 station network resolved in ~1 hour/cpu.

1,000,000 rinex files in 2 days on 44-cpu cluster
Progress on Analysis

1994-2007:

IGS

BARGEN

SCIGN

PANGA

EBRY

EUREF

NEARNET (240 station semi-continuous network)

PPP + ambizap takes 7 days on 44-cpu cluster
Mid-Term Prospects


UNR solution for SNARF

See preliminary solution by Kreemer et al.

Still need to carefully screen time series
Ambizap in GIPSY

In collaboration with JPL, ambizap is being
implemented in future distribution of GIPSY (where
ambizap follows PPP performed by “gd2p.pl”)

Can be implemented on ~4K cpu Caltech cluster

No practical limit to number of stations (e.g., could
easily be tens of thousands per day).
Long-Term Prospects

PPP does not improve orbits



So does not improve global-scale parameters
Implement ambizap into global IGS processing

Goal: one consistent global-scale solution

Orbit determination using ~1,000 stations, with
ambiguity resolution (carrier range)!
This in turn will improve PPP, and so on.

Possibly an iterative solution to this.

Preliminary scheme has been designed.

Will lead to improved reference frames.