A Multi-Objective Optimization Framework for Enhancing Traffic Efficiency and Safety in Intelligent Transportation Systems Using Real-Time Sensor Data and Predictive Analytics

• Aiman Hakimi

  Universiti Teknologi Sarawak, Department of Computer Science and Digital Systems, Jalan Tun Ahmad Zaidi, Sibu, Sarawak, Malaysia

  Author

• Danish Izzuddin

  Universiti Sultan Azlan Shah, Faculty of Computing and Multimedia, Jalan Raja Permaisuri Bainun, Kuala Kangsar, Perak, Malaysia

  Author

• Ariful Islam

  Dhaka, Bangladesh

  Author

• Firdaus Rahmat

  Universiti Malaysia Kelantan, Department of Information Systems, Jalan Pengkalan Chepa, Kota Bharu, Kelantan, Malaysia

  Author

Intelligent transportation systems increasingly rely on real-time sensing and analytics to manage congestion, reduce crashes, and coordinate multimodal flows under operational uncertainty. Modern urban networks generate large volumes of heterogeneous data from loop detectors, cameras, probe vehicles, and connected infrastructure, yet translating these streams into consistent control actions that scale across corridors and jurisdictions remains challenging. The gap is amplified by delayed or missing measurements, endogenous demand responses, and the need to reconcile multiple goals such as delay reduction, throughput preservation, emission mitigation, and safety risk attenuation. This paper studies a multi-objective optimization framework that fuses streaming sensor data with predictive models to adaptively allocate control resources, including signal phase splits, variable speed limits, ramp metering rates, and route guidance incentives. The framework integrates state estimation, risk-sensitive prediction, and constrained decision making within a unified structure that targets efficiency and safety without privileging one objective in a static way. It leverages learned surrogates for computational tractability while preserving interpretable constraints that align with engineering practice and regulatory bounds. The resulting architecture is designed for online use with bounded computation and explicit robustness to model mismatch and stochastic disturbances. We provide the conceptual elements, mathematical formulation, algorithmic building blocks, and an evaluation strategy using realistic streaming scenarios. The approach emphasizes neutrality with respect to technology choices and can interoperate with existing control hardware and center software, providing a pathway toward incrementally deployable improvements in traffic performance and safety outcomes.

• PDF

Copyright (c) 2025 authors

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