Digital-Twin-Enabled Remote Operation of Autonomous Systems: An Architecture-Driven Survey

Abstract

Digital twins are emerging as a key enabling technology for the remote operation of autonomous systems, offering enhanced situational awareness, predictive insight, and decision support through synchronized virtual representations of physical systems and their environments. However, the rapid growth of interconnected systems across diverse operational design domains poses significant challenges in timely perception, reliable communication, effective human oversight, and trustworthy decision making throughout the control loop. This article presents an architecture-driven survey that introduces a generic architecture connecting the physical twin, the communication network, the digital twin, and the operator within a single remote operation loop, around which the literature is synthesized. Based on this architecture, we classify how different types of digital copies support remote driving, remote assistance, and remote monitoring; synthesize the communication technologies, requirements, and achieved performance reported across domains; and examine the roles of artificial intelligence/machine learning, human–machine interaction, network digital twins, semantic communication, security, and dependability within the loop. The findings show that communication remains a major constraint, although end-to-end responsiveness is also affected by delays incurred by, e.g., sensing, processing, reconstruction, rendering, and mobility-induced events. At the same time, digital twins themselves can alleviate the attainable communication requirements by replacing bandwidth-intensive raw sensor streams with compact, synchronized state updates, revealing a reciprocal relationship in which the twin both depends on and relieves the communication system. The findings further show that faults and threats can propagate to other components across the remote operation loop, while the corresponding mitigation means in the literature remain largely developed and evaluated in isolation. To guide future work, we propose a minimum set of evaluation metrics and a validation-maturity roadmap, and identify semantic communication, network digital twins, quantitative safety assurance, and the transition from teledriving to telesupervision as the most promising directions toward safe and efficient digital-twin-enabled remote operation.