Abstract
<jats:p>Autonomous and heterogeneous marine robotic networks offer new possibilities for observing dynamic ocean phenomena beyond the limits of ship-centered and single-platform operations. Their scientific and operational value depends on coordinated system use: remote supervision, communication across constrained links, and onboard or near-real-time processing that converts raw sensor streams into compact information for live decision-making. This paper presents an evidence-based system framework for heterogeneous robotic operations in marine research coordinated through a Remote Operations Center (ROC). Zooplankton sampling is used as the motivating case because its patchy, dynamic biological structure is difficult to characterize efficiently with conventional ship-based approaches alone. The framework covers remote operation, edge processing and data quality, communication design, and targeted sampling with heterogeneous fleets. It is tested through a field case study conducted in Mausund, Norway, in 2024, where an autonomous underwater vehicle, two autonomous surface vehicles, and a support vessel were coordinated from a shore-based ROC approximately 100 km away. The results show that the ROC redistributed supervision, interpretation, and coordination from the field site to a shared shore-based environment, allowing these activities to continue in parallel with local launch, recovery, and support tasks. Compact edge-derived products made payload data more usable during operations, while layered communication design helped preserve command, coordination, and science-relevant information across links with varying capacity, latency, and reliability. The case study also showed how complementary platform roles supported broad-scale reconnaissance, local refinement, and targeted sampling. The findings indicate that the value of ROC-centered heterogeneous marine robotics lies in the combined observational capability of the system, linking platform roles, sensing, communication, coordination, and mission logic into one operational workflow.</jats:p>