Industries / Industrial equipment and services / Industrial machinery / Conveyors and material handling
Controls
Controls are the motors, sensors and software that decide when each stretch of conveyor runs, track every item, and tell sorters, storage machines and mobile robots what to do next.
5
Milestones achieved, from 1975 to 2019.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1975
Motors inside the rollers
Kazuo Itoh of Itoh Denki in Japan put the motor inside a conveyor roller in the mid-1970s, so that a conveyor could be built from short zones, each driven by its own roller and run only when a sensor sees an item in it. Itoh told Modern Materials Handling in 2010 that the idea caught on in Japan only after Panasonic fitted it in a VCR plant in the early 1980s, that DC and brushless DC motors from the mid-1980s made it a main conveyor drive, and that its use by the US Postal Service led equipment makers to adopt it in logistics. Itoh Denki dates its first brushless DC roller, the Power Moller 24, to 1986.
Credit
Kazuo Itoh, Itoh Denki
1998
A cargo terminal jams
When Hong Kong's new airport at Chek Lap Kok opened on 6 July 1998, the cargo handling system of SuperTerminal 1, run by Hong Kong Air Cargo Terminals Ltd (HACTL), ran so slowly that operators switched to manual mode, which could not carry the load and which they knew poorly. The Commission of Inquiry on the New Airport, reporting in January 1999, found that the probable main cause was poor integration between the terminal's control software, its Logistic Control System, and the machinery of its container and box storage systems, after building delays had squeezed the time for testing and training. HACTL had promised 75% of full capacity on opening day, and the commission found that a delay of about two months would have avoided the chaos.
LessonControl software has to be tested with the machines at full load, and staff drilled in manual mode, before opening day.
Credit
Commission of Inquiry on the New Airport (findings); HACTL
2008
Software closes the gaps
A sorter's controls need a gap between cases to track and divert each one, and on a shoe sorter that gap often sets the rate. Around 2008, Holste wrote, tracking software let shoe sorters run with 6-inch (15 cm) gaps instead of 12-inch (30 cm) ones, which raised them from about 120 to about 150 cases a minute without running faster. In 2017 he reported makers' figures that a PC-based gap upgrade could add up to 20% to an older sorter's rate.
Credit
Sorter makers, unnamed in the source, as reported by Cliff Holste in Supply Chain Digest
2019
One language for robot fleets
In 2019 the German Association of the Automotive Industry (VDA) and the machinery makers' association VDMA, with the Institute for Material Handling and Logistics at the Karlsruhe Institute of Technology, published VDA 5050, an open interface that lets one central fleet control system send orders to driverless vehicles from different makers. At a 2021 "AGV Mesh-Up", vehicles from six makers ran under a seventh maker's control system, and BMW's head of logistics planning for its Dingolfing and Debrecen plants said the interface was used in several BMW projects and set as the standard in its tenders for driverless transport systems. Version 2.0.0 was released in November 2021 and version 3.0.0 in March 2026, and at automatica 2025 robots from seven makers and institutes ran under one control system from MHP. A US alternative, the MassRobotics AMR Interoperability Standard, reached version 1.0 in May 2021; its public repository was last updated in November 2021.
Credit
VDA, VDMA and KIT IFL; BMW Group (user)
2019
Two warehouses at once
ASOS, the British online fashion retailer, overhauled the automation and technology in new warehouses in Atlanta and Berlin at the same time. In July 2019 it cut its profit forecast for the third time in eight months, saying the change had "taken longer than we had anticipated," hurting stock availability, sales and costs, and expected 50.5 million pounds of transition and restructuring costs; annual pretax profit fell 68% to 33.1 million pounds. In October chief executive Nick Beighton said the Berlin hub's automation issues were resolved.
LessonBeighton said that with hindsight ASOS had been "too ambitious in tackling two international warehouses at the same time.".
Credit
ASOS (chief executive Nick Beighton), as reported by Reuters
Watch for
Amazon's fleet traffic model
Amazon says DeepFleet, a generative AI model trained on movement data from its warehouse robots, coordinates the robots' routes and improves their travel time by 10%. A paper by Amazon researchers posted in August 2025 describes four designs trained on data from hundreds of thousands of robots; its abstract reports results on predicting robot movements, not the 10% figure.
Achieved whenAn outside party, or Amazon in data an outside reviewer can check, reports fleet travel time before and after DeepFleet at named sites.
Claimed by
Scott Dresser, vice president of Amazon Robotics, June 2025
Digital twin sets sort plans
A digital twin is a computer copy of a machine kept up to date with its live data. Researchers from Chinese universities working with SF Express report a digital twin of a cross-belt sorting system, used to choose which chute serves which destination, in use at 146 SF Express parcel centers in China, about 30% of the total. In field tests against SF's existing method they report average sorting throughput up 10.56%, peak throughput up 7.72% and parcels sent round the loop again down 60.81%.
Achieved whenA party outside the authors and SF Express repeats the gain, or SF Express reports it in its own operating figures.
Claimed by
Ran Li, Xiang Song and co-authors, Nature Communications, August 2026
Open
01
Cameras in place of photo-eyes
A short zone conveyor on which one overhead camera, instead of a photo-eye in each zone, sees every carton and tells each motor-driven roller zone when to run, compared on the same line with standard photo-eye control.
The briefCredit · unclaimed02
One test for conveyor control
A shared set of simulated conveyor layouts (a zone line with mixed carton lengths, a three-way merge, a shoe sorter with a minimum gap) with fixed traffic files and a scorer, so that control logic from any source can be compared.
The briefCredit · unclaimed
Open 01 · the brief
Cameras in place of photo-eyes
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A short zone conveyor on which one overhead camera, instead of a photo-eye in each zone, sees every carton and tells each motor-driven roller zone when to run, compared on the same line with standard photo-eye control.
Why it is still open
Zone conveyors start and stop each zone from a sensor that sees an item in it, as in Itoh Denki's design (1975, above). Cameras on conveyors so far watch rather than drive: Zebra sells a smart camera with AI software that tells a true jam from a false one, so that belts are not closed for false jams. Searches in November 2026 found camera jam monitors and camera tracking for robot picking, and no published system in which a camera releases the zones.
Built with what exists
Six 24-volt motor-driven rollers with their zone control cards (Itoh Denki or Interroll); a ceiling camera and a small computer such as an NVIDIA Jetson board; an open object detector such as YOLO, fine-tuned on a few hundred labeled frames; a microcontroller that switches each zone's run signal.
First step
Fit a camera over a three-zone roller line and log its detections beside the photo-eyes' signals for a day, before letting it drive any zone.
Done looks like
10,000 cartons, totes and polybags, at least a fifth of them black or clear, run under both kinds of control, with a published count for each of items touching, items missed, false stops and cartons a minute.
Open 02 · the brief
One test for conveyor control
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A shared set of simulated conveyor layouts (a zone line with mixed carton lengths, a three-way merge, a shoe sorter with a minimum gap) with fixed traffic files and a scorer, so that control logic from any source can be compared.
Why it is still open
Gains on sorters now come from software that controls gaps and merges, as Holste reported, but no public test compares such logic. The open-source Open Industry Project simulator models conveyors and runs its own PLC, yet its benchmark scripts time the simulator itself, and a GitHub search for a conveyor control benchmark found none.
Built with what exists
The Open Industry Project simulator and its built-in IEC 61131-3 PLC; Python for scoring; AI coding agents to write candidate logic (the Note "The jobs agents repeat").
First step
Rebuild the 2008 case: one shoe sorter fed at 12-inch and 6-inch gaps, and check that the rate rises as Holste describes.
Done looks like
Three published layouts with fixed traffic files, a scorer that reports cartons a minute, jams and gap errors, and results for at least three control programs, one of them written by an agent and one by a person.