Abstract Traffic signal inefficiencies in urban areas lead to increased travel time, fuel consumption, and delayed emergency response. This paper presents a modular, low-cost Smart Traffic Management System integrating infrared (IR) sensors, RFID overrides, and a supervised machine-learning model to dynamically control intersection signals. Four IR sensors detect vehicle presence at each lane’s stop line; readings are transmitted via USB–serial to a Python host that loads a pre-trained Decision Tree Classifier. The model predicts the optimal lane for the next green phase, and proportional green-time durations are computed based on real-time traffic density. An MFRC522 RFID reader enables immediate emergency-vehicle prioritization by overriding the normal cycle. Prolonged sensor activation (>15 s) triggers breakdown logging to a CSV file for maintenance alerting. A 16 × 2 I²C LCD provides real-time feedback—current green lane, countdown timer, and alerts. Laboratory testing achieved sub-100 ms decision latency and reduced under-utilized green-time by 30%. This framework delivers an affordable, extensible prototype for adaptive traffic control suitable for budget-constrained urban intersections.

| DOI: 10.17148/IARJSET.2025.125360