Back to search

National Technical University of Athens (NTUA)

Safety assessment and behaviour modelling of take-over dynamics in automated driving: from human response to network-level simulation

Abstract

dc:description

This dissertation aims to explain how take over response, driving performance and safety in automated driving at SAE Levels 2 and 3 is shaped by transition context, driver state, interface design and vehicle state dynamics, showing how their interaction during the transition affects safety at both the vehicle level and the network level. To address this aim, it brings together evidence synthesis, simulator experimentation, data-driven modelling, behavioural profiling and network-level analysis. This integrated framework makes it possible to examine the full chain of factors that influence the transition from automated to manual control and to link individual re-engagement behaviour with vehicle motion and mixed-traffic conditions. The systematic review and meta-analysis consolidate findings from SAE Levels 2 and 3 and show that time budget, scenario urgency, driver alertness, traffic conditions, road characteristics, demographic composition and Human-Machine Interface (HMI) characteristics shape re-engagement. These results also reveal substantial gaps in reporting, which limit comparability across studies and motivate the controlled and harmonised analyses that follow in the dissertation. The simulator experiment compares a baseline HMI with a multimodal innovative ΗΜΙ configuration and shows a clear shift in the transition pattern. Awareness time becomes shorter, take-over duration becomes longer and more deliberate, distraction decreases and interactions at small time to collision values occur less frequently, producing measurable improvements in objective safety indicators despite more differentiated subjective evaluations. Building on this, the data-driven modelling pipeline using XGBoost and SHAP quantifies how automation level, dynamic driving measurements, obstacle context, time budget and HMI design influence take-over duration. SAE Level 2 or 3 emerges as the strongest determinant of longer and more variable responses, while deceleration patterns, speed variability and scenario conditions further shape these responses. Behavioural profiling identifies three stable post-takeover driving profiles, passive-slow, nervous-moderate and normal-quick responders, which remain consistent across SAE Levels 2 and 3 and capture distinct modes of post-takeover control stabilisation. These behavioural parameters are then incorporated into a calibrated highway microscopic simulation model, completing the analytical chain. Spatial analysis of time to collision shows that take-over phases systematically reduce temporal safety margins, that lane closure and Operational Design Domain (ODD) exit scenarios create concentrated zones of elevated conflict potential and that merge areas and transition corridors intensify these effects. Higher market penetration of automated vehicles increases typical time to collision values but does not eliminate the bottlenecks associated with re-engagement dynamics. Taken together, these findings provide a compact and integrated account of how driver state, HMI support, behavioural variability and roadway context interact during take-over transitions, offering a consistent foundation for safer take-over management in future automated driving systems.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Sekadakis, Marios
  • Σεκαδάκης, Μάριος

Subjects

dc:subject × 18

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/60973

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Sekadakis, Marios; Σεκαδάκης, Μάριος. Safety assessment and behaviour modelling of take-over dynamics in automated driving: from human response to network-level simulation. National Technical University of Athens (NTUA), 2026. http://hdl.handle.net/10442/hedi/60973