NeuroSymbolic Traffic Reasoning with World-Grounded Multimodal Models for Explainable Autonomous Driving
Keywords:
neurosymbolic reasoning, world-grounded models, multimodal perception, explainable AI, autonomous driving, traffic governance, safety assurance, socio-technical infrastructureAbstract
The evolution of autonomous driving systems demands not only high perceptual accuracy and robust control but also a capacity for transparent reasoning that can be audited by human operators, regulators, and the public. This paper presents a comprehensive framework that integrates neurosymbolic reasoning with world-grounded multimodal models to achieve explainable decision-making in autonomous vehicles. We argue that purely end-to-end neural approaches, while powerful in perception, lack the compositional structure necessary for causal reasoning and accountability in safety-critical traffic scenarios. By combining neural perception modules with symbolic knowledge representations that capture traffic rules, social conventions, and environmental physics, the proposed architecture enables a system to derive interpretable explanations for its actions. Furthermore, the incorporation of world-grounded multimodal models—which align visual, linguistic, and spatial modalities with a shared semantic representation of the driving environment—enhances the system’s ability to reason about counterfactuals and hypothetical outcomes. This paper systematically examines the structural trade-offs inherent in such hybrid architectures, including the tension between neural flexibility and symbolic precision, the computational overhead of grounding mechanisms, and the governance challenges associated with certifying explainable behavior. We also explore deployment considerations such as real-time inference constraints, data heterogeneity across jurisdictions, and the need for standardized evaluation benchmarks. Through a cross-domain analysis that draws parallels with medical imaging, robotics, and infrastructure monitoring, we highlight the broader implications of neurosymbolic design for socio-technical systems. The paper concludes with policy recommendations for regulatory frameworks that mandate explainability in autonomous mobility, alongside a research agenda for sustainable, fair, and robust neurosymbolic driving systems.
References
1. Garcez, A. d. A., & Lamb, L. C. (2023). Neurosymbolic AI: The 3rd wave. Artificial Intelligence Review, 56(1), 1–34.
2. Marcus, G. (2020). The next decade in AI: Four steps towards robust artificial intelligence. arXiv preprint arXiv:2002.06177.
3. Mao, J., Niu, Y., Jiang, L., Bai, Y., & Zhang, Y. (2023). Multimodal learning for autonomous driving: A survey. IEEE Transactions on Intelligent Vehicles, 8(2), 1234–1250.
4. Li, Y., Li, S., & Savarese, S. (2022). Language-driven scene understanding for autonomous driving. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 1523–1532.
5. Manhães, M. S., & Ritt, M. (2021). Probabilistic logic programming for traffic rule compliance in autonomous driving. Journal of Artificial Intelligence Research, 70, 789–822.
6. van Krieken, E., Acar, E., & van Harmelen, F. (2022). Analyzing differentiable fuzzy logic operators. Artificial Intelligence, 302, 103602.
7. Harnad, S. (1990). The symbol grounding problem. Physica D: Nonlinear Phenomena, 42(1–3), 335–346.
8. Kim, J., & Crandall, D. (2024). Large language models for explainable decision-making in autonomous driving: A case study. ACM Transactions on Intelligent Systems and Technology, 15(3), 1–24.
9. Chen, L., Zhang, C., & Tao, D. (2023). Grounding language models in visual worlds: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(11), 12900–12920.
10. Pearl, J. (2009). Causality: Models, Reasoning, and Inference (2nd ed.). Cambridge University Press.
11. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., ... & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008.
12. Xiong, Z., Ye, X., Yaman, B., Cheng, S., Lu, Y., Luo, J., ... & Ren, L. (2026). UniDrive-WM: Unified Understanding, Planning and Generation World Model For Autonomous Driving. arXiv preprint arXiv:2601.04453.
13. Kim, B., Wattenberg, M., Gilmer, J., Cai, C., Wexler, J., & Viegas, F. (2018). Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (TCAV). Proceedings of the International Conference on Machine Learning, 2668–2677.
14. Rudin, C. (2019). Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nature Machine Intelligence, 1(5), 206–215.
15. European Commission. (2021). Proposal for a regulation laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). COM(2021) 206 final.
16. Awad, E., Dsouza, S., Kim, R., Schulz, J., Henrich, J., Shariff, A., ... & Rahwan, I. (2018). The moral machine experiment. Nature, 563(7729), 59–64.
17. Katz, G., Huang, D. A., Ibeling, D., Julian, K., Lazarus, C., Lim, R., ... & Barrett, C. (2023). Verification of neural network visual perception for autonomous driving. Formal Methods in System Design, 62, 1–28.
18. Dreossi, T., Donzé, A., & Seshia, S. A. (2019). Compositional falsification of cyber-physical systems with machine learning components. Journal of Automated Reasoning, 63(4), 1031–1054.
19. Sun, J., Cao, Y., Wang, W., & Zhang, Z. (2022). Real-time neurosymbolic reasoning for autonomous driving on embedded platforms. Proceedings of the IEEE International Conference on Intelligent Transportation Systems, 1–8.
20. Gerla, M., & Kleinrock, L. (2021). Vehicular networks and the future of the internet of vehicles. IEEE Communications Magazine, 59(6), 22–27.
21. Schwartz, R., Dodge, J., Smith, N. A., & Etzioni, O. (2020). Green AI. Communications of the ACM, 63(12), 54–63.
22. Michelmore, R., Wicker, M., Laurenti, L., Kwiatkowska, M., & Gal, Y. (2020). Uncertainty quantification for deep neural network-based autonomous driving systems. Proceedings of the AAAI Conference on Artificial Intelligence, 34(4), 5296–5303.
23. Van der Maaten, L., & Hinton, G. (2008). Visualizing data using t-SNE. Journal of Machine Learning Research, 9, 2579–2605.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Computational Intelligence Systems
This work is licensed under a Creative Commons Attribution 4.0 International License.
This article is published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
