AI AND IOT INTEGRATED SMART TRAFFIC SIGNAL MONITORING AND CONTROL SYSTEM

Abstract

Traffic congestion at urban intersections causes long waiting times, higher fuel consumption, increased air pollution, and delays for emergency vehicles. Most signals in developing cities still operate on fixed timing plans that do not change with the actual number of vehicles waiting on each approach. As a result, green time is often given to empty roads while long queues build up on busy ones. This paper presents an AI and IoT integrated smart traffic signal monitoring and control system that measures traffic conditions at an intersection in real time and adjusts signal timings according to the observed demand. The system uses cameras mounted at each approach of the intersection together with a Raspberry Pi edge controller that runs a lightweight object detection model. The model counts cars, two-wheelers, buses, trucks, and auto-rickshaws in each lane and estimates the length of the queue. Infrared sensors placed at stop lines serve as a backup when camera visibility is poor due to rain, fog, or glare. The detected counts are converted into a weighted traffic density for each approach, giving larger vehicles a higher weight because they occupy more road space and take longer to clear. An adaptive timing algorithm running on the controller allocates green time in proportion to the traffic density of each approach, within minimum and maximum limits that ensure fairness and pedestrian safety. The controller drives the signal lamps through a relay board connected to an Arduino or ESP32 module. When the vision model detects an ambulance or fire engine, or when a siren is detected by a sound sensor, the controller gives priority green to that approach so that the emergency vehicle can pass without stopping. Every intersection controller publishes vehicle counts, signal states, queue lengths, and fault conditions to a cloud platform using MQTT. A web dashboard shows the live state of all connected intersections on a city map, and traffic police can override timings remotely during events or accidents. Historical data is stored and used to analyse peak hours, identify recurring bottlenecks, and train a prediction model that estimates traffic volume for the next few minutes so that timings can be adjusted in advance. Simulation and small-scale testing with recorded traffic video showed that adaptive timing reduced average waiting time noticeably compared with a fixed-time plan, particularly when traffic was unevenly distributed across approaches. The system uses affordable hardware and can be installed on existing signal poles without major civil work. Future work includes coordination between neighbouring intersections to create green waves along arterial roads and detection of traffic violations such as red-light jumping.