Extreme deep-sea environments, such as hydrothermal vents and cold seeps, pose major challenges for ocean observation because active fluid discharge, steep chemical gradients, fragile ecosystems, and rugged terrain require measurements across multiple spatial and temporal scales. We report the development and field application of a 6,000-m-class heterogeneous underwater robotic fleet comprising an ROV, seafloor landers, and AUVs for coordinated operations in ocean science. The system has been applied to hydrothermal fields in the Okinawa Trough and cold-seep systems in the South China Sea, with the goal of linking fine-scale seafloor processes to broader water-column environmental patterns.

Within the fleet, the ROV serves as the high-precision intervention and close-range sensing platform. It performs visual inspection, targeted sampling of fluids, rocks, sediments, and biological communities, and deployment/recovery of in situ sensors and experimental modules in complex terrain such as hydrothermal chimneys, diffuse-flow zones, seepage patches, and biological habitats. The lander provides a stable seabed node for long-term observation and controllable experiments, supporting time-series measurements of physicochemical parameters, fluid/seepage activity, benthic response, and experimental incubations under in situ pressure, temperature, and hydrodynamic conditions. The AUV provides regional context through autonomous near-seafloor mapping, bathymetric survey, acoustic imaging, and water-column environmental profiling, enabling construction of three-dimensional fields of terrain, plume distribution, current-driven transport, and chemical anomalies.

By integrating these platforms, the fleet enables a workflow from wide-area autonomous reconnaissance, to target identification and precise ROV intervention, to long-duration lander-based monitoring and experimental verification. This coordinated mode improves sampling efficiency, reduces the uncertainty of single-platform observations, and provides multi-scale constraints for understanding material fluxes, ecosystem dynamics, and environmental impacts in deep-sea extreme environments. The results demonstrate that heterogeneous robotic fleets can serve not only as operational tools, but also as integrated ocean science infrastructure for process-oriented observation, adaptive sampling, and experimental studies in the deep ocean.