Real-time vehicle control via edge cloud sensor fusion and CNN based perceptron

Area of Science:

• Intelligent Transportation Systems
• Edge Computing
• Deep Learning

Background:

• Real-time vehicle control is critical for intelligent transport systems, relying on rapid sensor data processing for perception and decision-making.
• Current systems face challenges in adapting to varying environmental conditions and ensuring operational safety.

Purpose of the Study:

• To develop a hybrid edge-cloud method integrating deep learning with Internet of Things (IoT) sensor fusion for adaptive vehicle control.
• To enhance object detection, stopping-time prediction, and braking control under diverse driving scenarios.

Main Methods:

• Utilized ultrasonic range data fused with convolutional neural networks (CNNs) for perception tasks.
• Implemented a hybrid edge-cloud architecture for processing and control.
• Trained and evaluated the CNN model under normal and simulated adverse driving conditions on edge hardware (Jetson Nano, Raspberry Pi).

Main Results:

• Achieved high performance with R² = 0.99 (normal) and R² = 0.98 (adverse conditions).
• Recorded a low mean squared error (MSE) of 0.0085.
• Demonstrated low inference latency (110-230 ms) suitable for real-time edge deployment.

Conclusions:

• The hybrid edge-cloud approach enables adaptive, real-time vehicle control through IoT sensor fusion and CNN-based perception.
• The system enhances prediction accuracy and operational safety in variable driving conditions.
• Confirms the feasibility of deploying deep learning on low-cost edge devices for intelligent transport applications.