Development of a simulated annealing–adaptive neuro-fuzzy inference system (SA-ANFIS) model for sorghum seed intra-row spacing optimization
Year: 2026
Type: article
Abstract: Uneven intra-row seed placement remains a major challenge in precision sorghum planting, reducing stand uniformity and input-use efficiency. This study developed and evaluated a Simulated Annealing–optimized Adaptive Neuro-Fuzzy Inference System (SA-ANFIS) for predicting sorghum seed spacing under varying soil-moisture levels and planter forward speeds. Experimental data from an instrumented smart planter were screened, filtered, normalized, and partitioned using fixed procedures to improve transparency and minimize bias. The final dataset comprised 120 valid observations, using soil moisture and planter speed as inputs and intra-row spacing as the output. A first-order Sugeno ANFIS was first trained using hybrid learning, and its premise parameters were subsequently optimized using Simulated Annealing. Model performance was assessed using MSE, RMSE, R², Bland–Altman analysis, and response-surface evaluation. Optimization reduced normalized training MSE from 0.0028 to 0.0011. Test performance improved from RMSE = 0.8171 cm and R² = 0.830 for the baseline model to RMSE = 0.4595 cm and R² = 0.898 for the optimized model, representing a 43.8% reduction in prediction error. The optimized controller was successfully implemented on an Arduino Mega 2560 with a 50–100 ms control interval, demonstrating practical embedded feasibility. Although validation was limited to prototype testing, the SA-ANFIS framework provides a reliable and reproducible basis for intelligent sorghum seed-spacing control.
Language: English
Subfield: Mechanical Engineering
Field: Engineering
Domain: Physical Sciences
SDG: Zero hunger
Open Access status: diamond
OpenAlex ID: W7170970873