Autonomous and self-adaptive computing systems are rapidly becoming foundational components of contemporary digital infrastructure, ranging from cyber-physical systems and smart grids to intelligent software platforms and large-scale distributed architectures. This transformation is driven by advances in sensing, computation, connectivity, and control, yet it also introduces unprecedented challenges related to reliability, trustworthiness, architectural longevity, and rigorous system design. The increasing autonomy of systems reduces direct human oversight, thereby shifting the burden of correctness, safety, and dependability onto architectural principles, formal design methodologies, and runtime adaptation mechanisms. This article presents an in-depth, theory-driven research synthesis that integrates reliability theory, architectural innovation, dynamic reconfiguration, and rigorous system design frameworks to establish a cohesive understanding of how autonomous systems can be engineered to be dependable over extended lifetimes.

Drawing strictly on the provided body of literature, this work bridges traditionally fragmented research domains, including reliability engineering, computer architecture, formal methods, self-aware computing, and system sustainability. It examines how foundational reliability concepts inform the design of systems operating on unreliable hardware fabrics, how architectural strategies mitigate scaling-induced reliability degradation, and how formal models of interaction and reconfiguration enable predictable autonomy. Particular attention is paid to the role of world modeling, controller synthesis, and trustworthiness in autonomous behavior, emphasizing the necessity of aligning runtime decision-making with design-time guarantees.

The article further explores dynamic reconfigurable architectures as a response to environmental uncertainty, component degradation, and evolving system goals. By examining rigorous frameworks such as BIP-based modeling and distributed interaction optimization, the study demonstrates how autonomy can be systematically constrained to remain within safe and reliable operational envelopes. In parallel, perspectives from self-aware and self-expressive computing are incorporated to highlight how systems can reason about their own state, performance, and reliability, thereby enabling informed adaptation rather than reactive adjustment.

Through extensive theoretical elaboration, the article identifies critical gaps in current research, particularly the lack of unified methodologies that combine architectural reliability, formal verification, and adaptive intelligence. It argues that long-term sustainability and trust in autonomous systems depend not on isolated techniques but on the coherent integration of reliability theory, architectural rigor, and self-awareness. The contribution of this work lies in articulating a comprehensive conceptual framework that synthesizes these domains, offering a foundation for future research and system development aimed at building autonomous systems that are not only intelligent and adaptive, but also robust, trustworthy, and enduring.