DLG Field Robot Event: integrated autonomy is not yet farm deployment

External-source research synthesis, not an OfferSignal client result. Based on the retained October 2026 research and source-specific dates.

A farm robot can identify a plant and still fail if it cannot stay between rows, signal the finding, reach the right spot or recover safely. The 23rd DLG Field Robot Event tests that complete request-to-action loop in a maize field—while showing how far a competition result is from a buyer-ready product.

What the event actually tested

From 16–18 June 2026 at DLG Feldtage in Bernburg, Germany, 16 university teams from 10 countries and 135 participants competed with self-built autonomous robots. DLG describes three days and 13 hours of in-field competition. Team Acorn, associated with Osnabrück University of Applied Sciences and the University of Osnabrück, finished first; Robatic Bullseye of Wageningen University was second and FREDT of TU Braunschweig third. [1] [5]

The constructed test used maize rows, coloured disease markers, artificial insect stations and coloured soil spots. Most runs lasted three minutes. Tasks 1–4 equally determined the overall result: row navigation, unhealthy-plant detection, insect classification and soil-spot treatment. Task 5 freestyle was judged separately. Crop or marker damage was penalised, making this more than a software demonstration—but it was still a bounded event, not a multi-farm production trial. [2]

Three autonomy trade-offs

Acorn’s 45 × 90 × 64 cm, 22 kg prototype used four-wheel drive, a zero turning radius, more than 1.4 m/s stated speed, an Ouster lidar and two RealSense cameras. A Jetson Orin AGX ran Robot Operating System 2 nodes, state machines and behaviour trees with computer-vision and inference tools. Its stated strategy was to build a point-cloud view of rows and plants, generate a path, follow it and use depth cameras for in-row detection. [6]

The Osnabrück account adds that a second robot, Bcorn, carried a precision arm and drill for soil treatment; an arm-mounted depth camera positioned the drill and a 270-degree camera found markers. This is custom integration around a narrow task, not a catalogue product or documented commercial workflow. [3]

Robatic Bullseye was a 40 × 130 × 75 cm, 65 kg platform with four-wheel drive, a stated maximum speed of 6 m/s, 40 cm turning radius, two lidars and two ZED 2i stereo cameras. FREDT’s Helios Evo was 35 × 69 × 40 cm and 30 kg, with four-wheel drive, all-wheel Ackermann steering, 3.5 m/s stated speed, dual-height lidar and RealSense cameras. Its public design emphasised fast reaction and row geometry. These are team-provided specifications; no independent uptime or per-task score sheet is published. [7] [8]

What the results do—and do not—show

The event supports a concrete integration lesson: performance depended on sensor placement and calibration, row following, detection or classification, visible or digital signalling, physical action and repeated testing. But the public record does not publish winner-versus-runner-up score margins, raw telemetry, crop-damage counts, confusion matrices or run-by-run variance.

A participant report provides a useful counterexample. LCASTOR finished joint eighth in navigation, joint fifth in disease-marker detection after finding some artificial flags, and 13th in insect classification after false detections. Its strawberry-picking arm placed second in freestyle, but the arm was teleoperated from the arena edge. “AI detection” is therefore not the same as generalisable field performance or autonomous harvesting. [4]

An illustrative buyer-side test

For an illustrative agricultural-equipment manufacturer, start with one crop, row geometry and intervention. Version the sensor, software and task rules; preserve the detection log and physical action; and define who can approve treatment or stop the machine. Test wet soil, glare, occlusion, battery limits, recovery and crop damage—not only the curated fixture.

The authoritative decision remains with the agronomist, safety owner or farm operator, not the model. Measure repeatability, false actions, recovery time, uptime and agronomic effect separately. A robot that wins a three-minute task has demonstrated capability under those rules; it has not demonstrated price, serviceability, compliance, farmer willingness to pay, production deployment or return on investment.

The commercial boundary

This was a non-commercial, student-driven university competition supported by academic partners and DLG/Feldtage. The highlighted systems are prototypes. No public source gives purchase price, lease, per-hectare charge, service-level agreement, production availability, labour-hour reduction, yield effect or total cost of ownership. The 2027 event announcement signals continued experimentation, not 2026 commercialisation. [10]

The practical takeaway is disciplined rather than promotional: Acorn, Bullseye and FREDT are credible experiments in integrated agricultural autonomy. Their strongest lesson is that perception, geometry, controls, actuation and human-readable authority must work together. The evidence stops before buyer acceptance and routine farm operations.

Sources and next reading

Explore OfferSignal · Focused Business & AI Workshop · All research

References

[1]: https://www.dlg.org/en/detail/-field-robot-event-2026-winners-announced-international-FRE "Field Robot Event: 2026 winners announced" (DLG, 8 October 2026). [2]: https://onecdn.io/media/fre2026tasksv04-1-0d6a2387-d8d3-4267-b0a1-93fa2fab5dac.pdf "Field Robot Event 2026 – Task Descriptions, V0.4" (official/public event task document). [3]: https://www.hs-osnabrueck.de/nachrichten/2026/06/osnabrueck-wieder-weltmeister-in-feldrobotik/ "Osnabrück wieder Weltmeister in Feldrobotik" (Osnabrück University of Applied Sciences, 30 June 2026). [4]: https://lcas.lincoln.ac.uk/wp/lcastor-at-field-robot-event-2026/ "LCASTOR at Field Robot Event 2026" (University of Lincoln Lincoln Centre for Autonomous Systems, 29 June 2026). [5]: https://www.claas-foundation.com/en/foundation/news/2026-07-02-fre "Field Robot Event at the DLG Field Days" (CLAAS Foundation, 2 July 2026). [6]: https://www.fieldrobotevent.eu/team-acorn "Team Acorn" (official Field Robot Event team profile). [7]: https://www.fieldrobotevent.eu/team-robatic "Team Robatic Bullseye" (official Field Robot Event team profile). [8]: https://www.fieldrobotevent.eu/team-fredt "Team FREDT" (official Field Robot Event team profile). [9]: https://eurageng.eu/?tag=field-robot-event "Field Robot Event" (European Society of Agricultural Engineers topic page). [10]: https://www.fieldrobotevent.eu/ "FIELD ROBOT EVENT" (official public event portal; current page announces the June 2027 edition).