Traffiti

Inspiration

While driving on Sherbrooke Street, at the intersection, it came to our mind that transportation in general could be improved by attacking traffic light management.

What it does

Traffiti controls the traffic lights intelligently by using data on the current traffic to determine which traffic light pattern is the most appropriate to reduce overall wait time for drivers.

How we built it

We used Keras with Python on a traffic simulation software called SUMO to train the underlying algorithm that Traffiti is run on.

Challenges we ran into

The setup, learning the TraCI, which is the SUMO API, and familiarizing with reinforcement learning with Keras.

Accomplishments that we're proud of

Finishing the project on time and training the agent.

What's next for Traffiti

• Implement Traffiti IRL using computer vision
• Develop more features like protected lights, pedestrians, multi-lanes
• Store vehicle data for future usage on insurance, statistics, etc.

Built With

• keras
• python
• sumo
• traci