GPS data conditioning for enhancing reliability of automated off-road vehicles

Abstract

[EN] The practical implementation of precision agriculture at a large scale has not occurred yet for several reasons. Among them, the lack of uniformity and reliability in global positioning has discouraged many producers from adopting advanced solutions which, while considered to add a significant value to their production systems, cannot be incorporated without guarantees of minimum levels of long-term consistency. Although substantial improvements are constantly being introduced by receiver manufacturers, positioning errors can appear at the final stages of the localization process, resulting in inaccuracies and anomalies normally undetected by embedded quality filters. This article proposes an actuation protocol to enhance the robustness of GPS information for practical agricultural applications. The algorithm embodying this strategy merges partially-acquired raw strings into complete US National Marine Electronics Association messages whose information fields are checked for consistency. Once data qualifies as stable, other logic filters are applied to reinforce the likelihood of obtaining proper locations. Extensive field tests demonstrated that the algorithm was able to discard most erroneous positions due to typical GPS errors and poor signal reception in complex agricultural environments. However, the phenomena of coordinate quantization and random outliers were still present, which indicates that further redundancy is necessary to avoid unreliable outcomes. In this regard, positive results for supplementary consistency from GPS-based vehicle heading and speed are anticipated.