Seed Lab Automation: Hand Counts to ABB Robots

Before a new wheat or maize variety reaches a farm, thousands of seeds pass through a lab where people count them, sprout them, cut tiny samples from them and read their DNA. Seed lab automation is the effort to take that repetitive handling off human hands. This post looks at how seed testing and plant breeding labs worked in the past, how chipping and genotyping machines changed them, and what a Chinese wheat breeding robot shown in October 2026 says about the next step. It also covers where ABB lab robots such as GoFa and YuMi could fit.

A seed viability test at the USDA National Laboratory for Genetic Resources Preservation in Fort Collins, Colorado, 2016. Photo: Neil Palmer, USDA via Wikimedia Commons (public domain)

Seed counters, germination trays and breeders' notebooks

Seed testing began as consumer protection. Farmers in the nineteenth century often bought seed mixed with weeds, dust or dead grain, so in 1869 Friedrich Nobbe set up the first seed testing station at Tharandt in Saxony. Other countries followed, and in 1924 the International Seed Testing Association was founded to agree on common rules. The basic tests barely changed for a century: an analyst separated pure seed from rubbish by eye and with tweezers, then counted out seeds for a germination test. In spring, when farmers order seed, that meant long days of counting for the analysts on duty.

The Swedish Seed Testing Agency building at Stora Råby, which also housed the Nordic Gene Bank from 1979 to 1983. Photo: Dag Endresen / Wikimedia Commons (CC BY 3.0)

A standard germination test still uses 400 seeds, laid out in replicates of 100 on moist paper or sand and kept in a warm chamber. After a set number of days an analyst counts the normal seedlings, the abnormal ones and the seeds that never sprouted. Breeders did similar work at a larger scale. Norman Borlaug, who began breeding wheat in Mexico in 1944, grew two generations a year by shuttling seed between two regions, and his team threshed, sorted and recorded thousands of lines, much of it by hand.

Wheat breeder Norman Borlaug, photographed in 2004. Photo: Ben Zinner, USAID via Wikimedia Commons (public domain)

The first big automation came with DNA. Once breeders could link genetic markers to traits, they wanted to test a seed's DNA before planting it, without killing it. In the 2000s Monsanto developed automated seed chippers that held each seed, shaved off a tiny sliver for analysis and kept the rest alive in a tracked well. DNA extraction moved into 96 well plates handled by liquid handling robots. Counting got easier too, with optical seed counters and image based purity analysis taking over some of the work done with tweezers.

A technician handles seeds with tweezers under a microscope during a viability test at the USDA seed lab in Fort Collins. Photo: Neil Palmer, USDA via Wikimedia Commons (public domain)

Chipping, genotyping and a wheat breeding robot in 2026

In a large modern breeding lab, seeds are chipped, the chips are ground and sent through DNA extraction and genotyping, and the results decide which seeds go to the field. Germination and viability tests still involve a lot of hands, though. At the USDA's National Laboratory for Genetic Resources Preservation in Fort Collins, Colorado, staff check the viability of stored seed lots by placing seeds on blotters and in Petri dishes, one by one. Each sample then sits for days in a germination chamber before someone counts what has sprouted.

Seeds in a Petri dish for viability testing at the USDA National Laboratory for Genetic Resources Preservation. Photo: Neil Palmer, USDA via Wikimedia Commons (public domain)

The newest robot works in the field plots. On 10 October the South China Morning Post reported that Shandong Agricultural University had shown a wheat breeding robot at a breeding seminar. It uses high definition cameras, vision software and a robotic arm to measure wheat ears for traits such as grain count and ear length, and it was trained on images and expert scores from 2,000 different wheat lines. It marks promising plants with paint so breeders can look closer, and the university is working with Shandong Jizhi Biotech on a commercial version.

Demonstration wheat plots at the Cereals 2010 event in Cambridgeshire, England. Photo: Michael Trolove / Wikimedia Commons (CC BY-SA 2.0)

ABB has no seed lab customer that we could find, but its robots already work in similar labs. ABB has built automated lab workstations with GoFa CRB 15000 cobots for XtalPi, a drug discovery company, and the dual arm YuMi was designed for small parts and lab tasks such as handling tubes. In a seed lab, a GoFa could move plates between a chipper, a grinder and a DNA extractor, and a YuMi could place seeds on germination paper. Such a cell would usually be built and tested first in RobotStudio.

Automated lab workstations at XtalPi, built with ABB GoFa cobots. Photo: ABB

Toward breeding labs that run themselves

Breeding cycles are getting shorter. The speed breeding method published in 2018, which uses long days under LED lights, can give up to six generations a year for spring wheat. More generations mean more seed to chip, test and sow, and that adds pressure to automate the lab steps in between. The idea is close to the self driving research labs in chemistry, where robots run experiments and software picks the next one.

A seed germination testing chamber in the seed laboratory at the Placerville Nursery, California, 2006. Photo: USFS Region 5 / Wikimedia Commons (CC BY 2.0)

Germination testing is a likely next target. Cameras can already score seedlings in a tray, and a robot arm could lay out the 400 seeds, move trays in and out of chambers and photograph them each day. ABB's mobile YuMi concept, a dual arm robot on a wheeled base that loads centrifuges in hospital labs, shows the kind of machine that could serve several instruments in one room. For small seed companies, a shared cobot cell could be cheaper than hiring extra staff each spring.

ABB's mobile YuMi concept loading a centrifuge. Photo: ABB

Seeds vary in size, shape and stickiness, from round pea seeds to tiny grass seeds with awns, so grippers and suction cups have to cope with all of them. Labs need traceability for every seed, which means barcodes, cameras and software that never mix up two wells. Rules for official seed tests are written for human analysts, so a robot's result has to be accepted by regulators. The Shandong robot was trained on expert scores for exactly that reason: a breeder's eye for a good line is still hard to turn into data.

An ABB GoFa cobot working with lab instruments at the SLAS lab automation conference. Photo: ABB

Innovation

AI solutions for effortless ABB robot control.

Automation

Robotics

ceojohntran@chatwithrobot.net

+84905311611

© 2025. All rights reserved.

qtran1215@gmail.com