# Research basis and scope

## Published field evidence

The on-page research statement paraphrases DeBruin and colleagues, *Breaking the field phenotyping bottleneck in maize with autonomous robots*, **Communications Biology 8, 467 (2025)**. DOI: [10.1038/s42003-025-07890-7](https://doi.org/10.1038/s42003-025-07890-7).

The study reports multisensor collection from **198,249 experimental units** at **142 locations** during **2019–2023**, with derived phenotypic traits delivered for **132,176 experimental units**. These are different quantities: collected imagery does not automatically become a valid trait estimate. The research used EarthSense TerraSentia robots; it supports the practical relevance of under-canopy phenotyping, not the validation of this demonstration's three-robot coordination model.

Primary paper: <https://www.nature.com/articles/s42003-025-07890-7>

Open full-text mirror: <https://pmc.ncbi.nlm.nih.gov/articles/PMC11928528/>

## Compact robot scale

The University of Illinois TERRA-MEPP page lists an early TerraSentia model as **20 inches long × 13 inches wide × 13 inches high**. This corresponds to **50.8 × 33.0 × 33.0 cm**. It is a historical platform description, not a statement of current product specifications.

Official source: <https://terra-mepp.illinois.edu/about/process/robot>

The demonstrator uses an independently designed **30 cm wide × 50 cm long** robot, as a plausible design scale for the proposed virtual mission. It does not represent the commercial platform's dimensions, performance, software, hardware, or manufacturer endorsement.

## What the demonstration adds

The virtual fleet illustrates shared work allocation, data provenance, conflict management, and recovery from interruptions. These are design proposals for a browser simulation. The displayed trial is generated, not a replay of the published dataset. The paper's authors, EarthSense, the University of Illinois, and other research partners are not represented as partners in or endorsers of this Adaptive AgroTech concept.

The demonstrator should support questions such as: Which rows remain unmeasured? Which observations need review? How much mission time is spent waiting? What happens when one robot stops? Does the same mission benefit from additional robots? Such comparisons are useful for design discussion, but translating them into field performance requires experiments with real hardware and representative crops.

Sources checked on 12 October 2026 (UTC).