A Study on Deep Reinforcement Learning Based Traffic Signal Control for Mitigating Traffic Congestion

An edge-based situ-aware framework for energy-efficient 5G vehicular networks and traffic congestion mitigation

“…To address this problem, an edge-based situation-aware framework is proposed to reduce energy consumption and mitigate traffic congestion. The framework is composed of three (3) components: 1) vehicular energy saverturn off the unnecessary applications based on the LSTM model's forecast in driver's situation, 2) edge server load balancershare the loading of incoming requests to the neighbor servers and put the idle servers into sleep, and 3) traffic congestion mitigatorplacing realistic rewards on suggested rerouting road sections for reinforcement learning agents.…”

“…Intelligent or programmable equipment need to be purchased and additional man power has to be allocated. Fortunately, researchers have proposed more promising approaches based on machine learning [3]- [14]. Some studies use artificial intelligence (timeseries analytics) to learn traffic patterns in extending or shortening the timing on traffic lights [3]- [6].…”

“…Fortunately, researchers have proposed more promising approaches based on machine learning [3]- [14]. Some studies use artificial intelligence (timeseries analytics) to learn traffic patterns in extending or shortening the timing on traffic lights [3]- [6]. A study [14] proposes a novel approach to dynamically change the number of lanes in each direction based on analytics whereas a study [7] focus on rerouting vehicles away from congested areas, but shortest-path based algorithms are used which cannot reflect the everchanging road conditions.…”