This paper introduces an innovative and intelligent road navigation system designed to bridge the gap in autonomous capabilities for vehicles lacking Advanced Driver Assistance Systems (ADAS). The proposed method integrates machine learning techniques with real-time decision-making to provide a comprehensive solution for safer and smarter transportation. At its core is a YOLOv8-based object detection module capable of identifying potholes, traffic signs, symbols, vehicles, and other road objects in real-time.
The system uses CANBUS communication to ensure seamless integration between software modules and hardware components. CANBUS, a robust vehicle networking standard, enables efficient data exchange between the object detection module and Electronic Control Units (ECUs). These ECUs interpret detection signals to dynamically adjust the vehicle's speed, trajectory, and navigation path. The implementation employs STM32 microcontrollers coupled with CAN shields, where the STM32 handles real-time signal processing, and the CAN shield ensures high-speed, fault-tolerant communication across subsystems. This setup enables precise coordination of vehicle control modules.
The hardware design emphasizes modularity, allowing the system to be retrofitted into non-ADAS vehicles with minimal modifications. Additionally, sensor fusion techniques enhance the reliability and robustness of detections and decision-making under various environmental conditions. By extending autonomous navigation capabilities to non-ADAS vehicles offers a cost-effective solution for improving road safety, accessibility, and vehicle intelligence.