Reinforcement Learning based Dynamic Traffic Time Control and Enhancing Commercial Vehicle for Emergency situation with Smart Routing Techniques

This work proposes a Reinforcement Learning–based Dynamic Traffic Time Control system designed to create reliable, real‑time green corridors for emergency vehicles while simultaneously improving overall urban traffic flow. Unlike existing intelligent traffic management systems that primarily focus on congestion reduction through reinforcement learning and roadside unit (RSU) infrastructure, the proposed method introduces an integrated, emergency‑aware decision framework. Current systems often suffer from slow reaction to sudden disruptions, lack verification of emergency requests, and struggle to coordinate multi‑intersection signal control, leading to unreliable support for ambulances during critical moments. The proposed solution addresses these limitations through a three‑layered mechanism: (1) Emergency authentication, performed using GPS‑based movement patterns and hospital‑side confirmation to eliminate false alerts; (2) Short‑term traffic prediction using a neural network model to anticipate congestion trends before they form; and (3) Adaptive signal optimization using Proximal Policy Optimization (PPO) to dynamically allocate green times across intersections. A composite reward structure minimizes emergency‑vehicle delays while maintaining balanced traffic efficiency for non‑emergency users. The system is implemented on low‑cost Raspberry‑Pi IoT signal controllers communicating via secure MQTT/HTTPS protocols, evaluating latency constraints and failure mode contingencies under real-time constraints. Under SUMO‑based simulations on 4×4 and 10×10 grid configurations, alongside validation on highly irregular, real-world urban road network layouts, the proposed approach ensures dependable green‑corridor creation. Quantitatively, the system achieves a significant reduction in emergency vehicle delay and overall travel time compared to established fixed-time, actuated, and standard deep Q-network traffic control baselines, demonstrating viable scalability and responsiveness under unpredictable traffic conditions.