Adjustment of Real-time Kinematic-Global Positioning System (RTK-GPS) Survey Data: A Comparative Performance Analysis of the Back Propagation Artificial Neural Network (BPANN) and the Total Least Squares (TLS) Techniques

This study seeks to conduct an empirical evaluation of the performances of two soft computing methodologies comprising the Levenberg-Marquardt Back Propagation Artificial Neural Network (LMBPANN) and the Bayesian Regularisation Backpropagation Artificial Neural Network (BRBPANN). The study also assesses the performance of the soft computing techniques with the conventional Total Least Square (TLS) approach to calibrating Real-Time Kinematics Global Positioning System (RTK-GPS) survey data. The horizontal displacements (HD), arithmetic mean error (AME), arithmetic mean square error (AMSE), and arithmetic standard deviation (ASD) are the model evaluation and validation criteria used for the performance assessment. The analysis of results from the statistics viewpoint demonstrated that LMBPANN, BRBPANN, and TLS precisely adjusted RTK-GPS survey data with good precision in the study area. However, TLS better adjusts RTK-GPS survey data compared to LMBPANN and BRBPANN. Corresponding to the mean horizontal displacement measurement, the AME, AMSE, and ASD for TLS reached 1.41459E-09 m, 2.00428E-18 m, and 9.760E-14 m, and for the LMBPANN and BRBPANN, they reached 0.005595 m, 4.99277E-05 m, 0.000137 m, and 0.001287 m, 0.3.30633E-06 m, 4.06585E-05 m, correspondingly. The study concludes that although TLS is the most precise, BRBPANN offers a good alternative for adjusting RTK-GPS data in Ghana, thereby establishing a precise realistic technique for national and local applications.