Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{209798,
        author = {Sristi Yadav},
        title = {Graph Signal Processing for Traffic Sensor Placement and Network-Wide Traffic Recovery: A Survey and Case Study on METR-LA},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {5},
        pages = {3411-3426},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=209798},
        abstract = {Deciding where to place a limited number of traffic sensors, and how to rebuild speeds on unobserved roads from what those sensors report, is a long-standing problem in intelligent transportation systems. Classical sensor location models focus on flow observability and origin–destination estimation, while recent data-driven models forecast traffic with deep neural networks on road graphs. Graph signal processing (GSP) sits between these views: it treats traffic speeds as a signal on the road graph and offers sampling theorems, sampling-set selection algorithms and reconstruction methods with explicit guarantees. This paper surveys GSP-based sensor placement and network-wide traffic recovery. We review the sensing and data background, the foundations of graph sampling theory, experimental-design and submodular approaches to sensor selection, and the main families of recovery methods: bandlimited projection, Laplacian-regularized, kernel-based, sparse and learned. We organize the literature in a taxonomy and compare the families by assumptions, guarantees, computational cost and robustness. A case study on the 207-sensor METR-LA highway graph, using greedy E-optimal placement on synthetic bandlimited signals, illustrates how placement, bandlimit and recovery method interact, and shows that the best recovery method depends on graph topology. We close with open problems, namely validation on real non-stationary traffic, dynamic graphs, graph learning, sensor failures, heterogeneous sensing and standard benchmarks, and give practical guidelines for agencies.},
        keywords = {E-optimal design, Graph sampling, Graph signal processing, Intelligent transportation systems, Laplacian regularization, METR-LA, Sensor placement, Survey, Traffic state recovery},
        month = {October},
        }