The paper highlights some software improvements we implemented between LPS 9.0 and 9.2 SP1, as some of the initial processing was performed in LPS 9.0. The improvements specifically pertained to the LPS ATE (Automatic Terrain Extraction) module, where we added quality improvements in the correlator. The authors found the ATE results with LPS 9.2 SP1 to be acceptable, with errors general under 2m. It is also important to note that we made Adaptive ATE available for satellite sensor models with the recent release of LPS 9.3, so I suspect the accuracy could even be further improved.
Here is another ISPRS paper from 2007 outlining some advances in CARTOSAT-1 data processing. It also highlights the use of LPS for the terrain processing component of the workflow.
CARTOSAT-2 was launched in early 2007 and also features two panchromatic sensors, featuring a greatly improved resolution of under one meter. A spec sheet from ISRO is here. The wikipedia article makes some interesting statements about resolution (80cm) and pricing, but I haven't been able to verify these statements. This paper, also from the ISPRS conference in Beijing, has some great information on both satellites, as well as some further details on the terrain processing workflow. LPS was used as well in this paper, although it does not state the version number.
Note that in addition to using LPS ATE for automatic terrain extraction, the LPS Terrain Editor can be used to view the imagery in stereo and perform interactive terrain editing (from manual compilation to editing an automatically correlated surface).
Note that in addition to the CARTOSAT-1 and 2 sensors, there is also a CARTOSAT-2a sensor that is reserved for military usa.