Sarthak Bibave, Anand Jadhav, Abhishek Agale, Aditya Kalyankar

DOI: https://doi.org/10.5281/zenodo.24197615

Artificial Intelligence (AI) has demonstrated substantial potential for traffic signal optimisation, congestion prediction, and incident detection in simulation environments. However, a persistent gap separates algorithmic performance from dependable real-world operation, particularly in mixed-traffic urban networks where motorcycles, three-wheelers, pedestrians, and non-motorised vehicles share limited road space. This paper does not propose a new AI algorithm or claim experimental validation. Instead, it presents a critical review of the deployment-readiness gap in AI-based traffic management and proposes a six-dimensional framework for assessing whether a system should be permitted to operate on public roads. The framework integrates: (1) AI monitoring, prediction, and signal optimisation; (2) mixed-traffic and pedestrian safety; (3) sensor and communication reliability; (4) cybersecurity and privacy; (5) human supervision and safe fallback; and (6) measurable evaluation criteria. Each dimension is anchored in existing literature and governance standards, notably the NIST AI Risk Management Framework. The review finds that existing surveys acknowledge the simulation-to-deployment gap but do not operationalise deployment readiness as an assessment condition. By positioning the framework as a diagnostic tool complementary to Technology Readiness Level assessments, this paper provides city authorities, traffic engineers, and system developers with a structured basis for judging when an AI traffic management system is operationally safe, rather than merely algorithmically promising.

AI traffic management, deployment readiness, mixed traffic, pedestrian safety, NIST AI RMF, operational resilience, intelligent transportation systems.

AI traffic management research does not need only another survey cataloguing algorithms or simulation results. It also needs a structured answer to the question that existing surveys raise but do not resolve: under what conditions should an AI system be permitted to control real traffic, and how can those conditions be verified? This paper has proposed a six-dimensional deployment-readiness framework addressing AI performance, mixed-traffic and pedestrian safety, sensor reliability, cybersecurity and privacy, human supervision, and measurable evaluation. The framework is anchored in gaps identified in the reviewed literature and in NIST AI RMF assessment requirements. It is positioned as an operational diagnostic complementary to policy-level readiness assessments such as TPRM. The framework’s value lies in preventing deployment decisions based on simulation performance alone, with safety, security, and oversight treated as post-deployment concerns. In mixed-traffic urban networks, where pedestrians, cyclists, and informal transport share constrained road space, the consequences of such decisions can fall disproportionately on vulnerable road users. The question is not simply whether AI can improve traffic management, but whether the conditions for doing so safely have been established. The proposed framework is a step toward answering that question.

[1] Y. Wu, C. Luo, J. Wang, Y. Wang, Y. Jiang, and Z. Yao, “A survey of large language models for urban traffic signal control,” Expert Systems with Applications, 2026, doi: 10.1016/j.eswa.2026.129093. [Verify publication status, volume, and article number.]
[2] U. Madububambachu, R. Fatima, A. Sherif, and K. Khalil, “Security and privacy solutions in intelligent transportation systems: A survey,” Internet of Things, vol. 34, Art. no. 101812, 2025, doi: 10.1016/j.iot.2025.101812.
[3] M. R. Albous, A. Lacheret, and M. Stephens, “The governance of AI-enabled transport: Bridging Kuwait’s tech-policy gap vis-à-vis the UAE and Singapore,” Transport Policy, 2026. [Add volume, pages/article number, DOI, and verify bibliographic record.]
[4] National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1, U.S. Department of Commerce, Jan. 2023.
[5] “RL-based multi-intersection traffic signal control: A structured survey of trends and open deployment challenges,” IEEE Xplore, 2026. [Author names, conference/journal title, pages, DOI, and publication record were not supplied; complete and verify before submission.]
[6] S. Omefe, “Assessment of the effectiveness of adaptive traffic signal control systems in reducing pedestrian-vehicle conflicts at high-risk crossings,” GSC Advanced Research and Reviews, vol. 24, no. 1, pp. 140–153, 2025. [Verify title, author details, and publication record.]
[7] A. K. Garasiya, “A systematic review of traffic signal design methodologies at urban intersections under heterogeneous traffic conditions,” InSight Bulletin, vol. IV, no. 2, pp. 1–7, 2026. [Verify journal/publisher and publication record.]
[8] “Comparative review of hybrid AI approaches for traffic signal systems,” IEEE Xplore, 2026. [Author names, conference title, pages, DOI, and publication record were not supplied; complete and verify before submission.]

Sarthak Bibave, Anand Jadhav, Abhishek Agale, Aditya Kalyankar (2026). From Prediction to Safe Deployment: A Critical Review of AI-Based Traffic Management, Operational Resilience, and Governance in Mixed-Traffic Urban Networks. International Journal of Computer Techniques, 13(6), 756–759. ISSN: 2394-2231. DOI: https://doi.org/10.5281/zenodo.24197615