The ICONIC project was established to advance the way wind farms are operated by integrating intelligence at farm-, turbine- and component-level through cutting-edge modelling, control, and digital-twin technologies, enabling more efficient operation and improved power production. Within ICONIC, Work Package 2 (WP2) addresses farm-level decision-making and control. Its remit is to create an AI-driven framework that can coordinate all turbines in a wind farm. Deliverable D2.2, “A multi-agent, multi-layer, robust DRL-based wind-farm-control framework,” documents an important WP2 achievement: an intelligent wind-farm control strategy built on deep reinforcement learning (DRL).The detailed algorithms, training procedure, and simulation results have already been published in IEEE Transactions on Automation Science and Engineering. The paper is attached to this report. This report therefore serves as a cover note: it outlines the context, clarifies how the framework satisfies the objectives set out in the project plan, and directs readers to the published paper for full technical information. The aim is to give a clear overview of this deliverable while retaining complete traceability for anyone who wishes to examine the methodology and evidence in depth. Overall, D2.2 shows that the advanced AI concepts envisioned for ICONIC WP2 have been translated into a wind-farm-control solution, ready for the next stages of development and validation within the project.