Retrofittable ADAS for Heavy Vehicles

A Comprehensive Retrofittable ADAS for Heavy Vehicles Using YOLOv5 and Jetson Nano

Challenge

Heavy vehicles such as trucks, buses, and construction equipment face a critical safety issue: large blind spots on all four sides. Existing ADAS solutions are often expensive, vehicle-specific, or require OEM integration, making them inaccessible for older fleets and commercial operators. The absence of an affordable, retrofittable system means many vehicles still operate without the digital safety layer needed to prevent collisions during lane changes, turning, or reversing.

Our Solution

To address this gap, we independently developed a retrofittable Advanced Driver Assistance System that can be installed on any heavy vehicle without OEM dependency.

Key components of the system included:

• Real-time Object Detection:

  Implemented YOLOv5 models optimized for edge inference on the NVIDIA Jetson Nano, leveraging CUDA acceleration to maintain high FPS even with multiple camera inputs.

• Multi-Camera Image Stitching:

  Wide-angle cameras were combined to create a seamless, unified awareness system, enabling drivers to view blind spots from a single dashboard interface.

• Advanced Alert Systems:

  Context-aware warnings triggered whenever the system detected vehicles, pedestrians, or cyclists in critical zones. The alert logic was optimized to minimize false positives while ensuring driver responsiveness.

• Custom User Interface:

  A clean, intuitive UI displayed stitched video feeds, detection overlays, and alert states in real time.

• Remote Monitoring Module:

  Developed Android apps and WebSocket-based backend communication to allow fleet managers to remotely monitor detection events and live camera feeds.

This system was designed to be fully retrofittable, affordable, and scalable—making modern ADAS features available to vehicles that traditionally lack such safety technology.

Impact

The project demonstrated how low-cost edge AI hardware and open-source models can dramatically improve road safety for heavy vehicles. The system reduced blind-spot ambiguity, enhanced driver situational awareness, and offered fleet operators a data-driven approach to safety management. Its modular nature also makes it suitable for upgrades, including lane-departure modules, driver monitoring, or telematics integration.

Outcome

The work was presented orally at ICTIS 2025, Bangkok and has been accepted for publication in Springer’s LNNS series, pending guideline completion. The prototype validated the feasibility of deploying high-performance ADAS capabilities on compact hardware like Jetson Nano, opening pathways for future commercial development and further research.