Applications of machine learning in predicting rut depth in off-road environments

Area of Science:

• Agricultural Engineering
• Soil Mechanics
• Machine Learning Applications

Background:

• Rut depth from off-road vehicles significantly impacts vehicle performance and soil compaction.
• Conventional methods struggle with the complex, nonlinear relationships affecting rut depth, leading to estimation errors.

Purpose of the Study:

• To accurately predict rut depth caused by off-road vehicles using machine learning.
• To enhance prediction accuracy by integrating optimization algorithms with a machine learning model.

Main Methods:

• Utilized the Categorical Boosting (CatBoost) algorithm for rut depth prediction.
• Integrated Gray Wolf Optimization (GWO), Particle Swarm Optimization (PSO), and Secretary Bird Optimization Algorithm (SBOA) for hyperparameter tuning.
• Collected 270 experimental samples under controlled conditions, varying vertical load, speed, traction device, and number of passes.

Main Results:

• The SBOA-CatBoost hybrid model demonstrated superior performance.
• Achieved a Root Mean Square Error (RMSE) of 0.35 mm and a coefficient of determination (R²) of 0.97707.
• Exhibited a Mean Absolute Percentage Error (MAPE) of 1.2%, indicating high prediction accuracy.

Conclusions:

• The SBOA-CatBoost model offers a highly accurate method for predicting off-road vehicle rut depth.
• This approach can aid in improving vehicle performance and mitigating soil compaction.
• Optimized machine learning models provide significant advantages over conventional estimation techniques.