Data-driven and hybrid digital twin models critically rely on the quality and quantity of available data. In engineering fields like structural dynamics, acquiring useful data is often hindered by several challenges most of which are directly or indirectly due to expensive experiments and non-existent operation scenarios. These limitations in most cases necessitate assumptions about the available data being informative— contains relevant patterns of interest, accurate— characterised by low noise levels and close to the truth, and reliable— reproducible under the same circumstances, which are not always met [1]. Consequently, it is often the case that a model is just as good as the underlying data and therefore depends on the data quality and volume. In engineering systems, the quality of experimental and operational acquired data is determined by the testing strategies and monitoring instrumentation respectively which often translates partly to where on the structure to measure data. This study addresses these challenges by proposing a novel framework that leverages the adaptive architecture of reinforcement learning. Our agent-based framework optimizes sensor placement and configuration using an information-theoretic reward function. This approach integrates optimal design of experiments and monitoring instrumentation with data assimilation and decision-support tasks of digital twins [2]. The presented framework operates by simulating various sensor configurations within a virtual model of the structure. The reinforcement learning agent iteratively adjusts these configurations to maximize the information content of the acquired data. This iterative process ensures that the most informative data is collected, thereby improving the overall accuracy and reliability of the digital twin model. Our preliminary results demonstrate that this approach can ensure the continuous assimilation of relevant data by the digital twin at various stages of its life cycle which further decreases uncertainty in predictions. REFERENCES [1] Cicirello A. Physics-Enhanced Machine Learning: a position paper for dynamical systems investigations. arXiv preprint arXiv:2405.05987. 2024 May 8. [2] Willcox K, Segundo B. The role of computational science in digital twins. Nature Computational Science. 2024 Mar;4(3):147-9.