Search for a command to run...
Abstract Real-time structural interpretation is essential for optimizing directional drilling operations, particularly in unconventional wells where gamma-ray (GR) measurements are often the only subsurface data available. Current GR-based geosteering workflows rely heavily on manual interpretation and engineering judgment, leading to inconsistent outcomes and delayed decisions. This study proposes a fully closed-loop system that integrates an automated GR-based structural interpretation engine (Auto-GR) with a directional drilling automation platform to enable real-time structural models that directly support steering decisions and demonstrate autonomous, formation-aware well placement in complex geological settings. At the core of this workflow is the Auto-GR system, which transforms one-dimensional GR data into three-dimensional structural models. Building upon the True Stratigraphic Thickness (TST) method, Auto-GR incorporates several advancements, including automated initialization and calibration using Dynamic Time Warping (DTW) to align the starting trajectory with reference well data, and dynamic selection of optimal pattern-matching window sizes to enhance model robustness. Azimuthal gamma ray (AGR) images are integrated as structural constraints, where directional patterns such as smile or frown shapes restrict possible hinging directions and reduce interpretation uncertainty. The platform further introduces lane detection and forward projection capabilities: lane detection identifies structurally consistent drilling corridors based on GR-derived stratigraphy and historical well paths, guiding the trajectory toward geologically favorable regions, while forward projection extends the current structural trend to forecast likely formations ahead of the bit, enabling proactive steering updates. Together, these features create a continuous feedback loop between real-time subsurface interpretation and directional control, supporting formation-aware autonomous drilling analogous to self-driving vehicles adapting to changing road conditions. The integrated Auto-GR and DD Advisor system was validated through multiple unconventional horizontal drilling case studies. The real-time structural models generated by Auto-GR closely matched expert interpretations, and the lane detection function consistently identified formation boundaries that aligned with known productive intervals. Forward projection allowed early recognition of structural deviations and timely trajectory adjustments, while the incorporation of AGR image data significantly reduced interpretation uncertainty. Ensemble analyses confirmed a substantial reduction in model variability when AGR constraints were applied, demonstrating that the integrated system improves the accuracy, consistency, and responsiveness of geosteering operation while reducing dependency on manual adjustments and accelerating decision-making in real time. This study introduces a real-time closed-loop geosteering framework that links AI-driven GR structural interpretation directly to directional drilling control. Unlike traditional workflows that separate interpretation from execution, this system unifies model construction and steering logic within a single automated feedback loop. The combination of automated initialization, adaptive pattern matching, AGR-informed constraints, and lane detection provides a robust and scalable solution for navigation in subsurface environments with limited data. By embedding high-frequency interpretation updates directly into the control logic, this approach moves the industry closer to fully autonomous well placement guided by real-time subsurface intelligence.