Intelligent Tire Sensor-Based Real-Time Road Surface Classification Using an Artificial Neural Network.

Area of Science:

• Vehicle dynamics and control
• Artificial intelligence in transportation
• Sensor data fusion

Background:

• Advanced driver assistance systems (ADAS) and autonomous driving rely on accurate environmental perception.
• Estimating road surface conditions is crucial for optimizing vehicle control and safety.

Purpose of the Study:

• To develop and evaluate a real-time road surface classification algorithm using deep neural networks.
• To compare the performance of fully connected neural networks (FCNN) and convolutional neural networks (CNN) for this task.
• To identify optimal network configurations for accurate and timely road surface classification.

Main Methods:

• Collected acceleration data from an intelligent tire sensor system with a three-axis accelerometer.
• Trained FCNN and CNN models using longitudinal and vertical acceleration signals.
• Evaluated network performance based on classification accuracy and input data parameters.

Main Results:

• The CNN demonstrated sufficient real-time accuracy in classifying road surface types using longitudinal and vertical acceleration signals.
• A multi-input CNN capable of processing 2-axis or 3-axis acceleration data was proposed to enhance classification accuracy.
• Analysis revealed the impact of the number of classes and input data length on classification accuracy and delay.

Conclusions:

• The proposed deep learning-based algorithm enables real-time road surface classification.
• CNNs, particularly with multi-axis input, are effective for this task.
• The algorithm can be integrated into various vehicle electronic control systems to improve performance and safety.