Quick Answer: Remote crane operation turns a private 5G port network into an uplink system. In a field validation published in May 2026, Konecranes found that about 95% of a remote operation payload travels from crane to operator, and that uplink capacity, not container stacks, limits how many cranes a cell can carry. Demand peaks while people still operate cranes remotely, before full automation takes the video off the air. An uplink-weighted time division duplex (TDD) frame helps, but spectrum rules decide whether a port can run one, in Germany’s 3.7 to 3.8 GHz band, the EU’s new 3.8 to 4.2 GHz band or the US Citizens Broadband Radio Service (CBRS). Size from the operator-to-crane ratio and settle the frame before designing the radios.
About 95% of the traffic between a remotely operated yard crane and its operator flows up, from the crane to the operator’s desk, according to a Konecranes field validation published in May 2026. In the yard, that looks like a rubber-tired gantry (RTG) crane straddling a row of boxes, cameras trained on the spreader, the trolley and the truck lane, every feed streaming to a control room. Commands travel the other way, a trickle by comparison.
The network plan behind the crane was probably drawn for handheld terminals receiving work orders. Coverage is clean over every block, and with one remote crane on a pilot, nobody sees a problem.
The second crane goes remote and the video starts to break up. By the third, operators are asking why the picture freezes mid-lift.
Hamburg Built the Testbed. The Applications Come Later.
Hamburger Hafen und Logistik AG (HHLA) announced on July 9, 2026 that a private 5G campus network had been serving more than one square kilometer of its Container Terminal Altenwerder since the end of May. Deutsche Telekom and Ericsson built it in locally allocated spectrum under PROCON-5G, a project in the Federal Ministry of Transport’s Digital Testing Grounds in Ports program. HHLA project manager Michael Albers says it lets the terminal and its partners test applications under realistic conditions and put them into operation straight away.
Beyond connecting vehicles, sensors, mobile devices and IT systems, HHLA hasn’t named the applications that will run on it, which is common for a private 5G port network at this stage. The capacity plan still has to assume something, and the default comes from enterprise and consumer networks, where traffic mostly flows down to devices.
Felixstowe shows how quickly that default ages. Hutchison Ports’ private 4G network there, Three Group Solutions says, was extremely reliable at delivering work instructions to tablets in vehicle cabs. Its private 5G replacement across Felixstowe, Harwich and London Thamesport, completed in January 2026, was driven largely by autonomous trucks that need multiple live video feeds when a remote operator steps in.
AI pushes the same way. Much of what the Port of Tyne runs on its Ericsson private network involves cameras, from live container scanning to vehicle-mounted cameras feeding an AI engine that flags road defects, according to Ericsson’s December 2025 update. Public networks show uplink rising too, though Ericsson’s June 2026 Mobility Report names apps, user uploads and cloud storage as the main drivers so far: uplink grew faster than downlink at 43 of the 55 service providers it measured. Its modeling suggests additional AI traffic could push uplink in 2031 to three times its 2025 level or more.
Fifteen Megabits per Crane, Sixty Under Load
Some of the most detailed public data on remote crane traffic comes from Konecranes, which has tested private 5G with Nokia since 2022. After a test-yard phase in Hyvinkää, Finland, the two ran a two-month proof of concept in the third quarter of 2025 at a live European terminal, with one rail-mounted gantry (RMG) crane and five RTGs on a 5G Standalone network.
In normal operation, each crane averaged 15 Mbps of uplink. Load testing with virtual cranes pushed the network’s two radio units to 360 Mbps combined, about 60 Mbps per crane, and a single crane’s device could reach 120 Mbps. Konecranes treats ten active remote sessions per cell as a safe planning assumption where similar spectrum is available.
Radios are usually mounted above the stacks, so the steel boxes caused less trouble than uplink capacity, which Konecranes named as the dominant constraint. A coverage-first design, the usual starting point on a metal-heavy factory floor, puts the effort in the wrong place.
The frame configuration moved the numbers too. On the same 100 MHz channel, switching from a 3:7 to a 4:6 slot configuration, as Konecranes labels them, raised the single-cell uplink limit from about 180 Mbps to about 260 Mbps.
| Setup | Channel | Slot configuration | Single-cell uplink limit |
| Hyvinkää test yard | 40 MHz, n41 | 3:7 | About 80 Mbps |
| Live terminal, configuration 1 | 100 MHz, n77 | 3:7 | About 180 Mbps |
| Live terminal, configuration 2 | 100 MHz, n77 | 4:6 | About 260 Mbps |
Source: Konecranes, May 2026.
The paper is candid about weak spots. Average round-trip latency across the full system was about 24 ms against a 20 ms target, and handovers between cells caused a 160 ms latency peak and eight communication-related stops on one crane. Each crane carried two routers, one for PROFINET control traffic and one for video, so the two flows could be handled separately.
| PRO TIP: Size from the operator ratioBefore modeling any cell, get the average and peak uplink per remotely operated crane from the crane vendor, and the operator-to-crane ratio for each rollout phase from the terminal. Multiply average uplink per remote session by the cranes that can be under remote control at once, not by fleet size, then check that simultaneous peaks and block changes still fit inside the cell. Konecranes notes that a 1:1 ratio on a small fleet can load the network harder than 1:3 on a larger one. |
| Scoping a port network? Start from field data.PrivateLTEand5G covers private network deployments across ports, rail, airports and manufacturing, including the performance figures integrators design around. Subscribe to PrivateLTEand5G for new analysis and deployment coverage. |
Automation Lowers the Uplink. The Years Before It Don’t.
Full automation is the easier case for the network. Automated cranes work from their own sensors and control logic, so the high-definition video a human operator needs no longer crosses the air. Safety traffic still needs low latency, but the volume falls.
Load peaks earlier, while people are still in the loop. Continuous one-to-one remote operation is one of the hardest cases Konecranes describes, and a remote block change carries the full remote payload for as long as it lasts. Many terminals run manual, remote and automated equipment side by side for years. A network sized for the automated end state will fall short long before the terminal gets there.
The same logic narrows which machines need wireless at all:
• RTGs. Diesel units, still much of the installed base, run without fixed power or data links, and extending fiber across a live yard is costly, so wireless often becomes the remote operation path and the main uplink load. OPCSA in Las Palmas chose private 5G for its RTG data partly to avoid fiber civil works.
• Quay cranes. Cable reels can carry fiber, so wireless more often handles telemetry or serves as a backup, and a backup has to carry everything the remote station needs the moment it takes over.
• Autonomous trucks. They need continuous connectivity, plus live video whenever a remote operator intervenes.
• Cameras on vehicles and drones. They add uplink wherever they move, and the Port of Tyne already uses both.
AI raises a design question for every camera. Run the model on the crane and the radio carries detections and alerts; send the video to an edge server, even one on site, and the full stream rides the uplink. Where inference runs is far cheaper to settle on paper than after the cells are built.
The Frame You Want Isn’t Always the Frame You’re Allowed
An uplink-heavy frame solves the capacity problem if the neighbors allow it. When one TDD network transmits downlink while an adjacent one receives uplink, the base stations interfere, so regulators either require aligned frames or impose guard bands and tighter emission limits on networks that don’t align. Public networks mostly run downlink-heavy frames.
In Germany’s 3.7 to 3.8 GHz band, where most campus networks run, the Bundesnetzagentur (BNetzA) sets no blanket synchronization requirement. But strict out-of-block limits for unsynchronized and semi-synchronized networks apply by default, and guard bands come out of the local licensee’s own spectrum, including toward the national operators below 3.7 GHz. To qualify for the relaxed limits, a network needs enough separation from its neighbors, or an agreement with them that BNetzA sees before go-live.
The EU is opening more room with Implementing Decision 2025/2425, which harmonizes 3.8 to 4.2 GHz for low- and medium-power local networks, allows unsynchronized operation of medium-power base stations and requires member states to make the band available by September 30, 2026. One coordination option it lists for protecting public networks lets a local network depart from the operators’ frame only by turning downlink slots into uplink, the direction a port wants. Germany consulted on adding the band to its frequency plan in May 2026.
In the US, CBRS leaves the question largely to a voluntary industry specification rather than FCC rules. Under the OnGo Alliance coexistence specification, CBRS devices (CBSDs) in the same coexistence group that cannot agree on a common frame fall back to LTE-TDD configuration 1 or 2, or the NR equivalents. An uplink-heavier frame on General Authorized Access (GAA) spectrum needs the neighbors’ agreement, one reason pilot spectrum choices often fail to travel.
| Band and region | Synchronization rule | What it means for an uplink-heavy frame | Status |
| 3.7 to 3.8 GHz, Germany | No blanket requirement; strict limits when unsynchronized | Guard band from your own block, unless you have isolation or a neighbor agreement | In force |
| 3.8 to 4.2 GHz, EU | Unsynchronized allowed at medium power | The most room for an uplink-heavy frame, at power capped for local coverage | Due by September 30, 2026 |
| CBRS, US | Fallback to LTE-TDD configuration 1 or 2 within a coexistence group | Depends on neighbors agreeing | Voluntary industry specification |
| PRO TIP: Settle the frame before you plan the radiosDecide the TDD configuration while the spectrum is being licensed, because changing it later can mean reopening agreements with neighbors. In Germany, open neighbor talks while preparing the frequency application, since BNetzA expects agreements to be notified before go-live. On CBRS, ask your SAS provider who shares your connected set before promising any uplink figure. Then write the agreed frame into the design basis so the capacity model and the license describe the same network. |
Where Does This Leave You?
If you’re a systems integrator or vendor bidding for private 5G port work, the deliverable that wins is an uplink model, with the coverage map as supporting detail. It should show peak remote sessions per cell, the operator-to-crane ratio for each automation phase, where AI inference runs, and the frame the spectrum license permits. Put uplink per crane and handover behavior in the bid, weak spots included; the Konecranes paper is more convincing for listing its eight stops. Vendors with field data from uplink-heavy deployments have something most port content lacks, and the Executive Voice Program and Partner With Us put it in front of the terminal operators writing these RFPs.
If you run IT or operational technology (OT) at a terminal, collect three things before the RFP: uplink per crane from your crane and automation vendors, the operator-to-crane ratio for each phase, and a list of every machine without a fixed data connection. Then confirm your spectrum position, including whether Europe’s new 3.8 to 4.2 GHz band changes your options. A wireless self-audit will show where today’s network already falls short. None of this is quick, and the remote operation phase that strains the network most tends to run long.
Frequently Asked Questions
Can’t we run remote cranes on a public 5G network slice instead?
For telemetry, asset tracking and other traffic that can tolerate delay, often yes. The primary remote operation link is a different case. A slice reserves a share of the operator’s capacity, but it runs inside the operator’s TDD frame, which on most public networks favors downlink, so the uplink ceiling stays where the operator set it. The terminal also inherits the operator’s maintenance windows, software changes and core location, none of which it controls. For a link that can stop a crane when it falters, that loss of control usually outweighs the savings on infrastructure.
Won’t an uplink-heavy frame starve our handhelds and tablets?
Usually not, since work orders and telemetry are small. The cost shows up in bulk downloads such as firmware updates, which can be scheduled for quiet periods. Check the downlink budget anyway if you plan video to vehicle cabs or augmented reality for maintenance crews.
How far does one design scale across a large terminal?
Adding cells raises total capacity, but each new cell also adds handover boundaries, and handovers produced the worst latency spikes in the field data above. A crane that crosses more boundaries per shift sees more of them. A larger site is also more likely to border other networks, and an uplink-heavy frame has to be agreed with each of them. Treat published figures, such as the roughly four cells Konecranes found workable in a typical yard, as a starting point for your own validation rather than a ceiling.
Is Wi-Fi really ruled out for remote cranes?
For continuous remote operation across a yard, it’s a hard fit. Standard Wi-Fi roaming between access points can take several hundred milliseconds unless fast-roaming features are carefully tuned, which is long enough to trip the emergency stop logic on a remotely operated crane. Unlicensed spectrum also means sharing the air with equipment the terminal doesn’t control, from truck-mounted radios to neighboring terminals’ networks. Wi-Fi still has a place for fixed equipment and non-critical data, and it can keep running alongside private 5G during a migration.
