# AI for Urban Mobility: Machine Learning Thesis Defense
> Explore a thesis on optimizing urban traffic signals using Deep Q-Network (DQN) reinforcement learning to reduce commuter delays and CO2 emissions.

Tags: thesis-defense, machine-learning, smart-city, traffic-management, reinforcement-learning, urban-mobility, ai
## Optimizing Urban Mobility: A Machine Learning Approach
* Thesis defense by James A. Thorne, Dept. of Civil Engineering.

## Problem Statement
* Static traffic signal systems cause a 32% increase in commuter delays.
* Current pre-timed cycles fail to adapt to real-time fluctuations, increasing carbon emissions.

## Research Objectives
* Develop a Deep Q-Network (DQN) model for multi-intersection signal control.
* Quantify waiting time reductions vs. standard controllers.
* Evaluate feasibility for edge computing deployment.

## Methodology
* Utilizes the SUMO (Simulation of Urban MObility) platform.
* Conducted simulation on a 4x4 grid network over 1,500 episodes.
* Reward function penalized by vehicle halt-time and queue length.

## Results & Performance
* The Proposed AI Model reduced wait times by nearly 45% compared to static systems.
* Average wait time: 68s (Static) vs. 45s (Actuated) vs. 26s (Proposed AI).
* Maintained stable throughput at high vehicle densities (up to 1650 vehicles/hour) where baseline systems failed.

## Key Findings
* **Adaptive Capability:** Successfully identified and dissipated 'phantom jams'.
* **Efficiency:** 20% improvement in network capacity utilization.
* **Environmental Impact:** Potential 12% reduction in idling-related CO2 emissions.

## Limitations & Future Work
* Current simulation assumes 100% sensor accuracy and excludes pedestrian behavior.
* Future goals include incorporating V2I (Vehicle-to-Infrastructure) data and multi-modal transit prioritization.
---
This presentation was created with [Bobr AI](https://bobr.ai) — an AI presentation generator.