Environment Modeling and Safe Decision-Making in Autonomous Driving Scenarios
INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES · 05.03.2026 · 6 (1) · 80-90
DOI: 10.66238/ijrss.v6i1.94 ↗Kurz gefasst
Fachartikel in der Zeitschrift INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES, online erschienen am 05.03.2026. Verfasst von Noah Petrov und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The rapid advancement of autonomous driving technologies has fundamentally transformed the landscape of intelligent transportation systems. A critical challenge in deploying these systems in complex, real-world urban scenarios remains the seamless integration of robust environment modeling with mathematically guaranteed safe decision-making frameworks. This paper presents a comprehensive investigation into the architecture, algorithmic foundations, and empirical validation of a novel pipeline that bridges high-fidelity spatial representation with safety-critical control mechanisms. We propose a hybrid system that leverages dynamic occupancy grid mapping and semantic vectorization for environment modeling, coupled with a constrained Markov Decision Process formulation resolved via safe deep reinforcement learning. By integrating Control Barrier Functions directly into the action-selection phase of the reinforcement learning agent, the proposed framework explicitly prohibits mathematically unsafe state-action transitions, ensuring collision avoidance even during the exploratory phases of learning. The methodology is rigorously evaluated within the CARLA simulation environment across a multitude of highly stochastic urban driving scenarios, including unprotected left turns, sudden pedestrian incursions, and adversarial multi-agent interactions. Empirical results demonstrate that the proposed architecture significantly reduces collision rates compared to traditional rule-based and standard learning-based baselines while maintaining optimal traffic flow efficiency and passenger comfort. This work provides significant theoretical contributions to the domain of safe artificial intelligence and offers a scalable, pragmatic solution for the next generation of autonomous vehicle software stacks.
Zitieren
APA
Petrov, N., Vance, B., & Whitmore, C. (2026). Environment Modeling and Safe Decision-Making in Autonomous Driving Scenarios. INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES, 6(1), 80-90. https://doi.org/10.66238/ijrss.v6i1.94
Vancouver
Petrov N, Vance B, Whitmore C. Environment Modeling and Safe Decision-Making in Autonomous Driving Scenarios. INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES. 2026;6(1):80-90. doi:10.66238/ijrss.v6i1.94
BibTeX
@article{Petrov2026,
author = {Petrov, Noah and Vance, Beatrice and Whitmore, Charles},
title = {Environment Modeling and Safe Decision-Making in Autonomous Driving Scenarios},
journal = {INTERNATIONAL JOURNAL OF RESEARCH & SOCIAL SCIENCES},
year = {2026},
volume = {6},
number = {1},
pages = {80--90},
doi = {10.66238/ijrss.v6i1.94}
}
EUVIMED