Real-time control for multi-parametric data fusion and dynamic offset optimization in sensor-based variable rate nitrogen application

Abstract

Real-time sensor systems for variable rate nitrogen (N) application (VRNA) are an established technology nowadays but they have some shortcomings in terms of their capability to consider multiple parameters relevant for plant growth. Further, the abundantly lacking section control in centrifugal spreaders limits the accuracy of a sensor-based VRNA, especially in combination with the temporal and spatial offsets between sensing and fertilizer placement. Fuzzy inference systems were incorporated into a real-time control to numerically fuse the crop N uptake sensed by a real-time sensor system, as well as mapped soil electrical conductivity (ECa) data for the calculation of site-specific N dose rates (DR). A distinction of two subsections within the working width of a sensor-spreader system was made based on the ECa data. Further, by implementing a generic model, the control system agronomically optimized the rate control of a centrifugal spreader in order to compensate positional lags and technical latencies and minimize the spatial offset between DR determination and application in a dynamic manner. With field tests at different driving speed scenarios going partly beyond the usual operation conditions, the real-time control was verified. The differentiation of the sections has resulted in slight DR differences, whereas the control system has shown a high consistency in calculating the DRs and sending commands to the spreader in a coordinated manner. The level of spatial concordance between DR determination and application had a highly stochastic character. However, the deviation was never beyond 1.5 m and the percentage of deviations beyond 1 m reached a maximum of 2.3% among the different recorded datasets, which can be considered as a sufficient performance for practical needs.