Vehicle-to-Vehicle (V2V) Communication for Cooperative Obstacle Detection Using Embedded Vision

Abstract

In modern urban environments, road congestion and collision risks demand cooperative solutions that extend beyond the capabilities of individual vehicles. This paper presents a Vehicle-to-Vehicle (V2V) communication system that combines embedded machine learning with low-cost wireless modules to enable real-time obstacle detection and alert sharing. Each vehicle node uses a Raspberry Pi with a camera, running a TensorFlow Lite model to detect predefined objects (cow, cat, dog, elephant, person). Upon detection, simple commands are transmitted via UART to an ESP8266 module, which forwards these alerts wirelessly using the lightweight MQTT protocol to nearby vehicles, creating a cooperative awareness network. The system was prototyped and validated on a two-vehicle setup within a budget of approximately INR 14,400. Experimental results confirm reliable object detection (accuracy exceeding 80%) and low-latency message exchange (less than 150 ms), demonstrating the feasibility of affordable V2V cooperative perception for safety applications. The proposed system is cost-effective, modular, and scalable, making it highly suitable for intelligent transportation systems and future autonomous vehicle ecosystems.