The purpose of this article is to present a non-hybrid fully closed-loop controller for the Artificial Pancreas (AP) problem focused on long-term clinical trials and home-use applications. It includes physical activity (PA) and unannounced meals. The controller is based on a robust gain-scheduled algorithm with a Linear Parameter-Varying (LPV) structure. It takes into account the time-varying dynamics of the problem by adapting itself in real-time according to measured glucose levels, and allows online fine-tuning during tests and periodic evaluations without the need of a controller redesign. The proposed fully parameterized LPV control adds several features to our previous results, accounts for the main perturbations of the AP problem and simplifies its implementation. To help in the parameter fine-tuning, a methodology based on clinical information is proposed. In-silico tests show that the achieved performance is similar or better than our previous Automatic Regulation of Glucose (ARG) algorithm, tested in two clinical trials, with the addition of the features mentioned before.