Context Aware Dynamic Traffic Signal Optimization

TL;DR

This paper proposes a next-generation traffic signal control system that leverages artificial intelligence and context awareness, incorporating fairness alongside queue length to optimize traffic flow more equitably.

Contribution

It introduces a novel AI-based traffic control algorithm that considers fairness, improving upon traditional methods focused solely on queue length or waiting time.

Findings

Enhanced traffic flow efficiency demonstrated

Fairness metric improves equitable vehicle treatment

Outperforms traditional control algorithms

Abstract

Conventional urban traffic control systems have been based on historical traffic data. Later advancements made use of detectors, which enabled the gathering of real time traffic data, in order to reorganize and calibrate traffic signalization programs. Further evolvement provided the ability to forecast traffic conditions, in order to develop traffic signalization programs and strategies precomputed and applied at the most appropriate time frame for the optimal control of the current traffic conditions. We, propose the next generation of traffic control systems based on principles of Artificial Intelligence and Context Awareness. Most of the existing algorithms use average waiting time or length of the queue to assess an algorithms performance. However, a low average waiting time may come at the cost of delaying other vehicles indefinitely. In our algorithm, besides the vehicle queue, we use fairness also as an important performance metric to assess an algorithms performance.