The question is not signal presence, but whether control tolerates its variations

Removing a cable from a robotic cell can facilitate the movement of an arm, a mobile robot, or a rotating platform. However, a device connecting to Wi-Fi or 5G does not prove that the network can sustain the temporal behavior required by a machine. The decision depends on the task, where the controller is executed, and what happens if packets arrive late or are lost.

In automation, latency is not equivalent to download speed. To evaluate a control connection, one must observe delay variation and losses, in addition to how the system responds to an interruption. A low average can hide punctual delays capable of de-coordinating a sequence. The figure needs a measurement context: packet path, network load, traffic periodicity, and observed extremes.

It is also important to distinguish local control from teleoperation. A master's thesis from the Universitat Politècnica de València developed an experimental platform to remotely control a robotic arm via gestures and compared 5G and Wi-Fi 5 networks. The architecture influences what is measured: a connection carrying commands to a local system does not necessarily represent the same loop as a network coordinating movements of several robots.

What a latency target expressed in milliseconds means

The appropriate threshold is determined by the process, not the 'real-time' label. A coordinated movement order may have different requirements than a supervision video, an inspection, or a teleoperation with human interaction. Even in a single cell, messages between robots are not necessarily equivalent to cyclic exchanges between a controller and a servo. Each flow requires its own acceptance criterion.

The Ponte algorithm, developed by researchers from the Universidad Politécnica de Madrid (UPM) and the Universidad Carlos III de Madrid (UC3M), manages the timing and transmission frequencies of industrial devices over Wi-Fi. UPM's communication reports a latency lower than 8 milliseconds for the described control, reliability expressed as 99.99% in the experiments, and simultaneous service to 40 devices. These are reported results for the evaluated method, not a general guarantee for Wi-Fi nor a feature any router offers out of the box.

The latency figure needs precision to be comparable: whether it measures one way or round trip; between which points the stopwatch starts and ends; the size and periodicity of the packets; what traffic they share; and how extreme cases are counted. If that context is missing, an isolated number does not allow deciding if the system meets the control cycle of a specific machine.

The article associated with Ponte is titled Time-Sensitive IIoT Flows over Wi-Fi: a Network Calculus Approach and was published in IEEE Internet of Things Journal, with DOI 10.1109/JIOT.2025.3623878. EurekAlert classifies the method as simulation or modeling. This characterization requires caution when transferring results to a real installation: industrial network behavior depends on its configuration and environment, not just a scheduling algorithm.

Wi-Fi, 5G, and Ethernet solve different problems

Wi-Fi offers a wireless option based on infrastructure and can be useful when coverage is designed for the plant and the application tolerates measured variations. Ponte is relevant because it attempts to order flows through timing and transmission frequency programming. UPM indicates that the algorithm can be incorporated into new-generation Wi-Fi routers. This does not equate to an automatic feature of all equipment: it depends on the implementation and the evaluated system.

An academic work from the University of Oviedo studies the real-time coordination of two robotic arms via 5G and proposes comparing performance with Wi-Fi 6E. The abstract identifies synchronization and evaluation objectives, but does not allow extracting an isolated universal threshold to choose between those technologies. The case serves as an evaluation design, not a general classification of 5G versus Wi-Fi.

Industrial Ethernet is cabling. LAPP describes its use in different automation levels and distinguishes the physical medium from WLAN; it also explains that radio waves can diffract, reflect, or attenuate when encountering obstacles. In applications where mobility is priority, radio can eliminate physical mobile connections; where a stable and dedicated transmission medium is required, cable remains an alternative. It is advisable to compare complete systems, including their operating conditions, not just protocols or generations.

The choice is not reduced to Wi-Fi versus 5G. The decisive point is what task is assigned to the connection: telemetry, supervision, setpoint sending, or temporal coordination. A system can maintain the fast loop locally and use radio for higher-level setpoints; another may rely on a remote connection for more functions. Separating those responsibilities reduces confusion between speed, network latency, and total machine response.

What case studies contribute and what they do not allow concluding

The UPV project developed a preliminary remote control system for a robotic arm with a gesture recognition interface, a Raspberry Pi, and 5G and Wi-Fi 5 networks. The abstract describes tests evaluating latency, stability, and accuracy, and notes that the platform could be extrapolated to industrial applications. It is an academic prototype, not an operation certification for a production line.

The University of Oviedo work studies the real-time coordination of two robotic arms via 5G and proposes comparing performance with Wi-Fi 6E. Among its objectives are temporal synchronization in static and mobile scenarios, the development of a use case with coordinated robots, and the evaluation of their performance. The case does not by itself establish a transferable result to another factory, another robot, or a different load.

Ponte provides another type of evidence: the university reports latency, reliability, and device count figures, while the article record identifies simulation or modeling as the research method. The correct reading is not to discard the result nor turn it into a promise for any environment, but to keep it linked to the evaluated conditions. A published figure does not replace acceptance tests in the installation where it will be deployed.

The three examples illustrate different scenarios: remote control of an arm, robot coordination over cellular networks, and industrial traffic scheduling over Wi-Fi. They are not identical comparisons and do not necessarily share the same measurement point, load, network design, or success criterion. A universal winner cannot be deduced from results that answer different questions.

Qué aportan los estudios de caso y qué no permiten concluir El proyecto de la UPV desarrolló un sistema preliminar de control remoto de un brazo robótico con una interfaz de recono...
Editorial illustration by BytePreview

The plant changes radio behavior

An industrial cell contains structures and equipment that can change propagation conditions. LAPP describes how obstacles and surfaces can reflect, diffract, or attenuate WLAN waves, and notes that a wireless link is exposed to different conditions than a cabled medium. Therefore, a measurement at one point with the plant static is not enough to represent all production states. Radio must be evaluated in real geometry and operation.

Device density also matters: sensors, cameras, terminals, and robots can share network resources and generate traffic of different priority. The maximum announced speed does not by itself indicate what delay a control packet will have when the network is busy or how many packets will be lost. Measurements must be taken with a representative load, including non-critical flows that will coexist with control.

Movement adds coverage changes and possible transitions between access points. For a mobile robot, an acceptable connection at a fixed position may not be sufficient during a complete path. The design must consider which functions continue locally if communication is lost and what condition causes the machine to stop safely. The response to disconnection is part of the design, not a question resolved only with coverage.

Wireless connectivity can be part of a robotic installation, but its design and evaluation depend on the functions using the network and the plant's operating conditions.

How to evaluate a deployment before removing cables

Before migrating, it is advisable to classify flows by criticality and describe which functions depend on the wireless connection. In teleoperation, one must identify where the loop closes, what remains under local control, and how a safe stop is activated. In multi-robot coordination, one must define what synchronization the task requires and what behavior is accepted upon delay or loss. The architecture must fit the specific function.

A useful evaluation measures the latency distribution, not just its average, in addition to jitter, losses and their bursts, continuity during movement, maximum expected load, and coexistence with non-critical traffic. Tests must cover representative positions and trajectories, different production conditions, and load changes. The acceptance criterion must be fixed before the pilot to avoid a visual demonstration being confused with system validation.

As minimum checks to decide:

  • Define the control cycle, measurement points, and maximum tolerable delay for each flow.
  • Measure with the expected load and record percentiles, extremes, losses, and recovery, not just averages.
  • Test displacements and coverage changes in the real geometry of the cell.
  • Verify which local operation and safe stop remain available upon disconnection.
  • Compare total cost, maintenance, and infrastructure needs of Wi-Fi, private 5G, and cabling.

The practical conclusion is conditional: wireless communication can be suitable for certain robotic functions, but the technology does not by itself guarantee stable control. Ponte reports specific results for industrial flows over Wi-Fi; academic works describe platforms and evaluations with 5G and Wi-Fi. The decision must be based on the control cycle and plant measurements, with defined limits for each task and a safe response to failures.