Continuous tracking of GPS signals in dynamic scenarios pose a significant challenge for the design of the tracking loops. Optimising a design to suit a particular scenario may degrade its performance in other scenarios. For instance, increasing the carrier tracking loop bandwidth to receive dynamic signals will inadvertently affect the accuracy of the raw measurements. Therefore, in a stand-alone GPS receiver, a trade-off design is required to perform optimally in all the scenarios. External sensor integration with the GPS is considered as an alternative to improve upon this, and INS is the ideal choice as it is not only autonomous but also provides attitude at higher data rates.

Traditionally, the integration of GPS and INS were carried out in loosely and tightly-coupled configurations. While these systems offer significant advantages over the stand-alone GPS, nevertheless, it is imperative to improve the performance wherever possible. Ultra-tight integration, also called "deep level tracking", integrates the I (in-phase) and Q (quadrature) signals from the tracking loops of the GPS receiver with the position and velocity obtained from the INS. The primary advantage of this configuration, in addition to the benefits of loosely and tightly-coupled systems, is a significant reduction of the tracking loops bandwidth, as the Doppler signal derived from INS aids the tracking loop to remove the dynamics from the GPS signals. Two important advantages stem from the reduction of the bandwidth: accuracy of the raw measurements, and increased immunity to jamming signals. The block diagram representing the 3 architectures are shown in Figure 1.

Figure 1: Block Diagram of Loose, Tight and Ultra-Tight Integration System

Unlike the loosely and tightly-coupled systems, which are considered to be 'feed forward', ultra-tight systems are 'feedback' systems, i.e., a feedback signal in the form of Doppler derived from the INS also drives the tracking loops. For the successful implementation of this system, the accuracy of the Doppler signal is very critical. The errors that degrade the Doppler accuracy are the residual biases in the estimates of integration Kalman filter. It is well known that inertial sensors suffer from systematic and stochastic biases, which degrade the navigation solution. Though the systematic bias is effectively removed by the integration Kalman filter through incorporation of GPS measurements, nevertheless, the stochastic component needs a different treatment. Appropriate modelling techniques, like Gauss Markov and Autoregressive, effectively mitigate this noise.

Conventional tracking loops obtain their feedback from within the channel. The Costas phase discriminator, which is the central part of the carrier tracking loop, generates the corrections from I and Q measurements. These corrections, after filtering, drive the NCO (Numerically Controlled Oscillator), which generates the quadrature signals to reduce the error with the incoming signals. This configuration is suitable for a system with low dynamics. However, as dynamics increase, the phase error transcends the threshold thereby loosing lock. Therefore, in the case of ultra-tight receivers, the NCO gets its correction signal not only from within the channel, but also from the INS. This additional signal from the INS removes the dynamics from the GPS signal and subsequently keeps the loop in lock. The feedback from the INS is the key to the ultra-tight receiverÕs performance. Figure 2 shows the ultra-tight tracking loop architecture.

Figure 2: Ultra-tight tracking loop architecture

Though the inertial aiding of the tracking loops seems to be attractive, if the aiding signal does not properly represent the true Doppler, it results in a tracking loop bias, which degrades the phase output of the Costas discriminator. This phenomenon, which is unseen in stand-alone systems, is prevalent in ultra-tight receivers or receivers with Doppler aiding. Ultimately, the challenge lies in removing any of the undesired effects created by the aiding signal.

Tracking Loop Performance. The tracking performance of the ultra-tight loop is compared with a conventional loop. The conventional loops have a limitation on the dynamics to be handled, and when the dynamics exceed the bandwidth, they switch back to wideband PLL or narrowband FLL (frequency lock loop) temporarily relinquishing the measurements for the navigation algorithm. This situation is aggravated in urban areas, indoors, etc. However, ultra-tight loops remain in the narrowband PLL mode even during high dynamics. Figure 3 shows a plot comparing the conventional and ultra-tight tracking loops for a constant velocity trajectory. While the conventional loops increase in frequency with the input, the ultra-tight loops with a bandwidth of only 3Hz maintain almost a constant Doppler.

Figure 3: Ultra-tight (BW = 3Hz) and Conventional (BW = 13Hz) tracking loops performance

The feedback Doppler signal removes most of the dynamics from GPS signals rendering the signal to be ÔalmostÕ dynamic-free. So, any residual dynamics on GPS signals to be tracked is only due to the local oscillator. As the Doppler due to oscillator is much less, carrier bandwidth can be reduced significantly improving the accuracy of raw measurements and anti-jam performance of the receiver.

Some preliminary research results were presented in:

• BABU, R., 2004. Mitigating the correlations in INS-aided GPS tracking loop measurements: A Kalman filter based approach. 17th Int. Tech. Meeting of the Satellite Division of the U.S. Institute of Navigation, Long Beach, California, 21-24 September, 1566-1574. (Download PDF)
• BABU, R., & WANG, J., 2004. Improving the quality of IMU-derived Doppler estimates for ultra-tight GPS/INS integration. GNSS2004, Rotterdam, The Netherlands, 16-19 May, CD-ROM proc., paper 144. (Download PDF)