AI-enabled Exit Strategy of Emergency Vehicle Preemption

• Main Authors: Khoirunnisaa, Kunti, Hartanto, Rudy, Wayan Mustika, I, Woraratpanya, Kuntpong, Arva Arshella, Ika
• Format: Conference Proceeding
• Language: English
• Subjects: Artificial intelligence; Complexity theory; Electrical engineering; EVP; Exit Strategy; Law enforcement; Real-time systems; Reinforcement learning; Safety
• ISSN: 2766-0419
• DOI: 10.1109/ICITEE59582.2023.10317753

Emergency Vehicle Preemption (EVP) is a system employed by traffic lights to prioritize emergency vehicles, such as ambulances, fire engines, and police cars. This technology allows these vehicles to safely pass through crowded intersections by stopping traffic from other directions. However, it is essential to consider that preempting the regular traffic signal control could have implications for the safety and efficiency of both the prioritized emergency vehicle and the overall flow of regular traffic at the intersection. Research on EVP and Exit Strategy has been extensively developed and implemented in various commercial products, including the Trafficware 980 ATC V76 signal controller. Trafficware offers multiple EVP schemes and Exit Strategy modes that can be customized by traffic operators to suit their specific requirements. Studies have indicated that certain Exit Strategies perform optimally under specific traffic conditions. Yet, the complexity of traffic conditions poses a challenge for traffic operators to optimize the Trafficware's performance by selecting the most effective Exit Strategy mode. To address this, Artificial Intelligence (AI) with Reinforcement Learning (RL) techniques used to improve the Trafficware 980 ATC V76 system. Particularly, the Exit to Fixed Phase strategy will be optimized, enabling Trafficware to dynamically choose the most suitable exit phase from the available four phases based on real-time traffic conditions. Implementing this optimization will replace the current fixed exit phase selection with a more adaptable and responsive approach, leading to improved traffic management and efficiency, ultimately resulting in reduced waiting times. The results show that the AI-enabled exit strategy can reduce the waiting time with the best performance up to 7.8% compared to no exit strategy. The Trafficware 980 ATC V76 has been upgraded with the ability to adjust signals based on policies to recover from congestion caused by preemption.