In urban areas with narrow streets and tall buildings, none-line-of-sight (NLOS) or multipath (MP) signals severely attenuate the positioning accuracy of the global navigation satellite system (GNSS). vector tracking loop (VT) has been employed in NLOS/MP signal detection and correction; its feasibility and superior performance have been assessed and demonstrated in previous research studies. In VT, the navigation solutions are fed back to calculate the signal tacking parameters and complete the carrier and code tracking. The NLOS/MP reception induces code delay to the code tracking. The generated local signal replicas are mixed with the received signal; the code and carrier discriminators’ outputs are employed as the measurements to estimate the navigation solutions in the VT navigation filter. In order to extract the code delay induced by the NLOS or MP, multiple correlators (MCs) are usually utilised to search for the code delay. However, more correlators will bring much heavier computation load and a real position is necessary for calculating the signal tracking parameters. The strap-down inertial navigation system (SINS) can generate short-term accurate navigation solutions. In this work, a gradient descend (GD) method is designed for extracting the NLOS/MP-induced delay in the correlators’ measurements level. The VT/SINS ultra-tightly coupled navigation system (VT-UTC) is developed to obtain mutual aiding between these individual systems. In VT-UTC, the corrected SINS navigation solutions are employed for calculating the signal tracking parameters, and reliable navigation solutions from the SINS can help to extract the code delays induced by the NLOS. Two field tests with global positioning system L1 signal are carried out for assessing the performance of the GD method in the VT-UTC framework. Comparisons between the MC and GD method show that the correlators’ computation reduces a lot since much fewer correlators are calculated in the code delay searching. Also, reliable navigation solutions from the SINS could help to calculate the signal tracking parameter while in-view GNSS satellites are limited.