<p>Dependability is a critical requirement for the effective deployment of robots in dynamic, unknown, and unstructured environments, such as the autonomous mining robots studied in the ROBOMINERS project. To address this challenge, we propose an application-agnostic framework that enables self-adaptation to unforeseen circumstances. This article explores the use of formalized models as a powerful tool for developers to enhance the autonomy and resilience of robotic systems. The core idea is to leverage these models at run-time to dynamically select the most suitable reconfiguration strategy in response to unexpected events. A key aspect of this approach is to ensure that the system continues to provide the expected service or benefit to users, even in the face of unpredictable events. Rather than relying on predefined or ad-hoc solutions, our adaptation strategies are computed during operation, using explicit engineering knowledge established during the design phase. This methodology relies on a meta-model that integrates principles from Category Theory—an abstract mathematical framework for expressing complex relationships—and Model-Based Systems Engineering. The meta-model is implemented using web ontologies, providing the flexibility to support diverse applications. During operation, the meta-model is used by a meta-controller, a high-level decision-making mechanism designed to achieve mission objectives, similar to how classical controllers regulate set-point references. To illustrate the effectiveness of this approach, we present three use cases involving mining robots. These scenarios address challenges such as sensor failures, insufficient capabilities (e.g. inadequate force application), and communication loss, showing the versatility and robustness of the proposed solution.</p>