Human-Machine Interface Dynamics and Driver Trust Calibration in Semi-Autonomous Vehicles Utilizing Real-Time OverFeat Perception-Based Lane Keeping and Collision Warnings

• Abey city

  Lautech

  Author

Semi-autonomous vehicle adoption faces a critical challenge: the discrepancy between system capabilities and driver trust calibration, manifesting as misuse (over-trust leading to complacency) or disuse (distrust leading to system rejection). Existing human-machine interface (HMI) designs inadequately address the dynamic nature of trust formation during real-world driving, particularly regarding perception system transparency. This study develops and validates a novel HMI framework integrating real-time OverFeat-based perception outputs (lane boundaries and obstacle detection) with trust-calibration mechanisms in semi-autonomous vehicles. Using a driving simulator experiment with 60 participants across varying automation levels, we measured trust calibration through behavioral (takeover response time, NDRT engagement) and self-report metrics. The OverFeat CNN architecture achieved 89.4% lane detection accuracy and 87.2% vehicle detection precision at >10 Hz processing speeds. Results demonstrate that real-time perception visualization significantly improves trust calibration (p < .01) compared to conventional black-box HMI designs, reducing takeover response time variability by 34%. The framework provides actionable guidelines for HMI designers and contributes to the theoretical understanding of trust dynamics in human-automation interaction.

• pdf

Copyright (c) 2026 Abey city (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.