Trusted Garden is designed as a sample room for a reproducible experimental farm. Cameras, soil probes, robotic scouts, water-quality instruments, edge computers and actuators are not hidden behind a dashboard—they are placed into a visible chain from observation to decision to verified action.
V0.8 is a refined design study for a buildable reference farm with a physical weather layer, higher-detail field hardware, and an instrumented control room. The 3D hardware layer is research-informed; operating values shown in the interface remain simulated until connected to physical sensors.
Paired-depth soil probes, weather context, RGB / thermal / multispectral imaging, flow meters, water-quality probes and acoustic–visual pond sensing create a synchronized state estimate.
A field gateway and pavilion edge racks handle local logging, model inference, device health and policy checks. RTK/GNSS and the site network keep observations spatially and temporally aligned.
Motorized valves, pumps, aeration, feeding and circulation are exposed as bounded actuators. Low-risk actions may execute automatically; higher-impact changes cross a human gate.
Robotic scouting, fixed cameras and water sensors measure the biological response after an intervention, closing the loop and leaving a Trust Trace instead of a one-way recommendation.
Treatment groups, biomass response and rhizosphere measurements become the experimental logic for the three field plots.
Audio + vision feeding-intensity modeling becomes a live sensing layer for the pond, paired with water-quality control.
The FishFreshNet line informs edge deployment, visual explanation and the idea that a research model should become an accessible tool.
Affordability, usability and human oversight become constraints on automation rather than afterthoughts.
Field rows use an approximately 0.76 m reference spacing, a common 30-inch research and production spacing cited by University of Minnesota Extension. Plant density is visually decimated for browser performance.
Each plot shows representative paired-depth soil sensing rather than a single “smart stick,” reflecting extension guidance to monitor more than one root-zone depth and more than one representative field location.
The east greenhouse combines a round fish system, aeration, solids separation, biofiltration, sump and deep-water-culture beds—an intentionally legible loop derived from FAO small-scale aquaponics guidance.
A straight public/research spine, transparent growing spaces and framed views borrow planning lessons from agricultural research centers where education routes are integrated with working production and research areas.
USDA-ARS projects describe autonomous field robots that combine GPS/IMU, LiDAR, computer vision and modular sensor payloads, as well as multimodal field phenotyping and UAV sensing. V0.5 turns those research patterns into a visible reference-farm equipment layer.
USDA-ARS has reported edge-computing video analysis for UAV crop monitoring, while commercial precision systems use camera-to-processor-to-actuator loops for site-specific operations. The model therefore shows field compute, machine-vision nodes and explicit actuator endpoints rather than a generic “AI cloud.”