Artificial Intelligence Rationale for Autonomous Vehicle Agents Behaviour in Driving Simulation Environment

Charissis, Vassilis; Papanastasiou, Stylianos · 2008 · OpenAlex-citations

DOI: 10.5772/5820

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of evaluating a novel automotive Head-Up Display (HUD) interface designed to improve driver spatial awareness and response times under low-visibility conditions. The authors argue that while HUD technology has advanced, user-centered interface design remains immature, often failing to effectively present sensor data without distracting the driver. To assess the proposed HUD’s effectiveness against traditional instrumentation, the researchers developed a cost-effective, open-source driving simulator. A critical component of this evaluation was the creation of realistic artificial intelligence (AI) for autonomous vehicle agents, ensuring that simulation scenarios accurately mirrored real-world traffic dynamics and accident-prone situations. The methodology involved modifying the open-source "The Open Racing Car Simulator" (TORCS) platform to create a custom simulator. The hardware setup included off-the-shelf components, such as a Logitech steering wheel with force feedback and a custom-built PC, housed in a laboratory equipped with recording devices. The software was reprogrammed to adhere to the British Highway Code rather than racing rules. The AI development combined macroscopic and microscopic simulation approaches. Macroscopically, vehicles were clustered into groups like "Stop-aheads" and "Jammies" to manage traffic flow complexity. Microscopically, individual agents were programmed with specific driving profiles and constraints. To enhance realism and immersion, the AI was designed to mimic human errors, such as delayed braking, while adhering to "hard constraints" (e.g., traffic direction) and allowing deviations from "soft constraints" (e.g., speed limits) to simulate accident scenarios. The study implemented two specific accident scenarios based on statistical data from the Strathclyde Police Department: a sudden braking event and a sharp turn with traffic congestion, both simulated under heavy fog conditions with visibility below 50 meters. The AI agents were programmed to create these hazardous situations, forcing human participants to react instantly. The simulator handled approximately twenty vehicles, using a looped track to create the illusion of continuous traffic. The AI rationale focused on generic reaction patterns derived from common sense driving behaviors, discounting irrational actions to reduce development time while maintaining sufficient unpredictability to challenge the driver. The significance of this work lies in its demonstration of a viable, low-cost method for conducting rigorous automotive HMI research. By leveraging open-source software and carefully calibrated AI agents that replicate human error and traffic flow, the authors created a flexible environment for measuring driver performance. This approach allows for the detailed analysis of how HUD interfaces impact response times in critical, low-visibility scenarios, providing a foundation for improving in-vehicle notification systems that prioritize safety over technological display.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Methodological Resource: tool software
• Theoretical Contribution: computational model