Integrating Edge Computing and Machine Learning for Low-Latency Decision Making in Next-Generation Intelligent Transportation Infrastructures

• Sujan Koirala

  Kantipur Engineering College, Department of Electronics and Telecommunication Engineering, Dhapakhel Road, Lalitpur, Nepal

  Author

• Niraj Shrestha

  Advanced College of Engineering and Management, Department of Communication and Network Engineering, Kalanki–Sitapaila Road, Kathmandu, Nepal

  Author

• Bibek Adhikari

  Himalaya College of Engineering, Department of Telecommunication and Information Systems, Chyasal–Imadol Road, Patan, Nepal

  Author

• Ariful Islam

  Dhaka, Bangladesh

  Author

Intelligent transportation infrastructures are expanding in scope and complexity as urban mobility evolves toward connected, automated, and electrified modalities. Continuous perception, prediction, and control across vehicles and roadside units increasingly depend on low-latency computation and robust learning under fluctuating wireless conditions. Edge computing provides proximity and bandwidth locality that can reduce round-trip delays compared to cloud-only designs, while machine learning models enable context-aware decision making across sensing, prediction, and resource orchestration. This paper examines an integrated design that combines edge computation with learning-driven scheduling to support latency-sensitive tasks that underpin safety and traffic efficiency. The focus is on architectural choices that couple distributed inference with cooperative control, on optimization formulations that codify latency and reliability targets, and on algorithmic mechanisms that preserve privacy and security at the network’s edge. A modular blueprint is articulated, with attention to the interfaces among sensing pipelines, model execution, data life-cycle management, and control loops. The study emphasizes rigorous delay accounting, stability of learning-enabled controllers, and mechanisms that guard against domain shift and adversarial behavior without inflating inference cost. Through an evaluation framework covering congestion estimation, collision-risk prediction, and adaptive signal timing, the paper assesses how placement, offloading, and prioritization affect end-to-end performance. The analysis aims to clarify tradeoffs inherent in balancing accuracy, responsiveness, and resource usage in next-generation deployments that must operate at scale under urban dynamics, thereby informing practical engineering decisions for resilient low-latency operation.

• PDF

Copyright (c) 2025 authors

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.