AI Controlled Traffic Management System
- 0 Collaborators
In this project, we developed a reinforcement learning model which given any road network and any traffic pattern (traffic densities which exhibit a daily, weekly or special routine) learns to optimize the flow of the traffic such that the clearance time of the traffic is brought down. ...learn more
Project status: Under Development
Overview / Usage
SUMO software has been used to generate random traffic in any random road network. The random traffic contains 1000 cars travelling from any sources to any destinations within the road network. Sensors at each road intersections are used to come up with traffic patterns which then becomes an environment for the algorithm to learn to resolve within the least time by controlling the traffic lights.
The project was presented at TCS EngiNX 2018 and won the 3rd prize in the contest.
Methodology / Approach
First, sensors at the road intersection determine the number of cars waiting at any red light. All information is stored to come up with a time-series dataset which establishes a pattern in the traffic. A reinforcement learning algorithm runs on the traffic pattern to solve it quicker, i.e. to clear all the vehicles from the random network as quick as possible.