Simulated Vehicle Pose Road Grade Recognition Method Based on Multilayer Perceptron

To address the road grade recognition problem in intelligent vehicle suspension control, a multilayer perceptron-based method was proposed. An improved harmonic superposition method was used to construct a 3D random road surface model. A seven-degree-of-freedom vehicle dynamics model was established via Lagrange equation to simulate vehicles’ vertical displacement, roll angle and pitch angle multidimensional dynamic responses. Based on these responses’ statistical characteristics, an improved fully connected neural network architecture was designed and trained. The method achieves a classification accuracy of 97.17% on the validation set and 85.74% on the generalization test. It adapts to 8-level road recognition under 10~30 m/s constant speed and 1 500~1 680 kg vehicle weight, providing direct road grade evaluation for intelligent suspension control.