In this research, we propose a system for monitoring and efficiently managing traffic congestion at the intersection of Park Avenue and E 72nd Street, New York, USA. This methodology is applied to a realistic traffic scenario where all intersections are controlled by fixed traffic signals. The method is based on a data-driven approach deploying a swarm of drones to measure the number of vehicles on roads. The collected information by the drones is sent to traffic lights and simultaneous perturbation stochastic approximation (SPSA) method is used to minimize traffic congestion by adapting the green traffic light durations. More precisely, we simulate the scenario thanks to the Vissim traffic software and Python. The simulation results highlight the effects of traffic light optimization to reduce traffic jams in contrast to the baseline case, where the duration of green lights is fixed.