Designing a Conceptual AI-Powered Warning Prototype to Enhance Safety During Overtaking

Sammanfattning

Overtaking is one of the riskiest maneuvers on two-lane roads, particularly in low- and middle-income countries where infrastructure and enforcement are limited. Existing Advanced Driver Assistance Systems (ADAS) provide reactive support such as lane keeping and collision warnings but rarely address the specific, predictive needs of overtaking. This thesis conceptually investigates how an Artificial Intelligence (AI)-powered, context-sensitive warning prototype could enhance overtaking safety in Iran, where road and behavioral challenges make this maneuver especially hazardous. The study employed the Design Science Research (DSR) methodology, integrating qualitative interviews with Iranian drivers and a participatory Future Workshop with traffic and IT experts. Thematic analysis revealed five central requirements: contextual awareness, personalization, multimodal alerts, explainability, and adaptability. These insights informed the design of a non-functional, Figma-based prototype that demonstrates conceptually how diverse data inputs (vehicle telemetry, Global Positioning System (GPS), radar, Light Detection and Ranging (LiDAR), environmental data, and driver state) could be fused into a predictive support system. The prototype includes a Trip Context screen for situational awareness, a Warning Head-Up Display (HUD) with explainable alerts, post-drive Analysis and Feedback screens for reflection, and a Settings module for personalization. The prototype has not been technically implemented or tested with real datasets; rather, it functions as a conceptual boundary object between research and practice. It visualizes how user-centered design and contextual adaptation can address limitations of current ADAS while laying the groundwork for future development. This work therefore represents a feasibility-oriented, design-level contribution rather than a fully developed application. The study contributes to literature by applying DSR in a novel safety-critical domain, by addressing the neglected issue of overtaking safety, and by introducing empirical insights from an underrepresented geographic context. It also contributes to practice by presenting a replicable conceptual design process, ethical design principles, and a blueprint for potential real-world implementation in resource-constrained settings. Future research should extend this work by collecting overtaking-specific datasets, implementing and testing predictive accuracy through field trials, and addressing ethical and regulatory requirements for safety-critical AI. By combining socio-technical insights with participatory design methods, the study demonstrates the feasibility and conceptual potential of intelligent, adaptive, and explainable overtaking-assistance systems.