Driving automation promises substantial public value, including more resilient public transport under labour constraints, improved accessibility, efficiency gains in freight logistics, and ultimately traffic safety benefits. Realising these benefits at scale is not solely a vehicle-technology challenge. It requires governance-capable operationalisation across the coupled domains of transport demand, vehicle development, legal approval, and infrastructure provision. In particular, the authorisation and continuation of driverless operation in defined operational domains hinges on the consistency condition that actual operational conditions remain within the described operational domain and within the system’s operational design domain (OC ⊆ OD ⊆ ODD). This turns the road space from a largely compliance-maintained asset into an operationally managed system with measurable service levels, trigger logics, and auditable evidence chains. This dissertation addresses a central element in that system: road markings are both safety-relevant guidance elements and a life-cycle-driven asset, yet existing quality measures are primarily human-centred and only weakly connected to machine perception. The central claim is that Machine Detectability (MD) provides the missing bridging construct that enables road marking quality to be specified, assessed, and managed in a manner compatible with automated driving and operational domain management. The dissertation operationalises MD through white-box metrics derived from raw camera, LiDAR, and radar data and links these measures to established infrastructure-side variables, including retro reflectivity (RL), wet retro reflectivity (RL,w), daytime visibility (Qd), and marking material and surface characteristics. Empirical results from field measurement campaigns and controlled test-field experiments quantify how MD relates to photometric properties across day and night as well as dry, moist, and wet conditions, and identify where linear predictors are reliable and where non-linearities, saturation, and wet-state effects constrain transferability. Methodologically, the work introduces a multi-parameter MD metric set that extends beyond sole contrast to include gradient and edge-related camera measures and LiDAR intensity/contrast representations and embeds these metrics in a level-based decomposition separating feature extraction and decision logic. Building on this structure, the dissertation proposes and utilises a modelling-and-inversion approach to derive context-dependent minimum requirements for RL, RL,w, and Qd for specified operational needs, including uncertainty via prediction intervals. Among other findings, the resulting analysis indicates that wet-night performance and RL,w are typically the most critical constraints in the prevailing operational status quo. In addition, the dissertation develops and validates a radar-detectable road marking concept using low-profile passive reflectors, demonstrating radar performance and durability under mechanical loading as a feasible pathway to sensor redundancy in adverse conditions. Overall, the dissertation establishes a coherent framework linking infrastructure measurement, maintenance management, and operational domain governance, and derives implications for road operators, standards bodies, OEMs, and approval authorities. Future research priorities include evidence-based MD thresholds linked to false-positive detections, scalable measurement and forecasting, wet-robust marking system design, tighter physical-digital alignment processes for temporary states, extension of MD to other road guidance elements, and KPI-based governance and contracting models to operationalise Automated Trafficability of roads.
Stefan Biermeier (Thu,) studied this question.