ESFSim will utilize its distributed, multi-modal observational and computational network to record measurements in both time and space, which will be used to forecast its own behavior, inform decisions, and enact changes through AI techniques. While agricultural technology is evolving rapidly in terms of irrigation systems, remote imagery, and mobile robotic platforms, the challenge remains to efficiently collect, process, and manage highly heterogeneous data streams from a smart farm in a manner that supports AI advancement benefiting farmers, workers, consumers, and the environment. The proposed ESFSim infrastructure is symbiotic with the campus smart farm being created at UC Merced. Some examples include ground-level sensor data, remotely-sensed data, images, and data collected from public sources, such as weather.

Data Processing, Preparation: Because real-world data is rarely useful in its raw form, processing will be conducted on the edge devices, as well as on the central service, so that the simulator makes use of the data. More specifically, this includes taking raw readings from sensors and organizing them into formats compatible with simulator models.

Data Access: Data will be accessible within the web interface provided for use with the dashboard, AI training tasks, and simulation tasks. Users will not be required to manually manage these datasets and will instead be able to interact with them through a documented interface.