Human-Machine Interface


The AI bottleneck: why sensing, not compute, defines the physical frontier


By Julien Ghaye, product line manager robotics at Melexis A


rtificial intelligence is most often discussed in terms of models, compute architectures, and the scale at which data can be processed. This framing


has driven rapid advances in decision-making capability, enabling systems to interpret patterns, predict outcomes, and optimize behaviour across increasingly complex domains. It is also reflected in the headlines that dominate the industry, where progress is most often associated with the latest model architectures, accelerator technologies, and the companies, like Nvidia, DeepSeek, OpenAI, Google, and Microsoft that develop them. Yet as intelligent systems move beyond the digital domain into the physical world, the assumption that the primary limitation lies largely in how data is processed, rather than how it is acquired, no longer holds. As these systems transition from purely digital environments into the physical world, they require more than intelligence alone. Like biological systems, they rely on continuous sensing to remain connected to reality.


This imbalance highlights a key requirement.


Intelligence, whether biological or


artificial, emerges from closed-loop interaction between sensing, decision and response. In this context,


sensing is not a peripheral component of AI systems but the foundation that enables the next-generation of solutions to function in the real world.


Understanding sensing as the basis of intelligent systems


At a system level, intelligent behaviour can be understood as a continuous loop: sensing the environment, interpreting that information, acting upon it, and observing the result. While recent innovation has focused heavily on interpretation and decision-making, the


30 May 2026


integrity of the loop depends equally on the quality of the initial measurement and the responsiveness of the feedback path. Sensors provide the critical interface between physical phenomena, human input, and digital processing. Whether measuring position, current, magnetic fields, or user interaction, they define how accurately a system can perceive both its internal state and the external environment it must respond to. This distinction is critical, as sensing underpins everything from closed-loop control of physical systems such as robotics to real-time feedback in human-machine interfaces (HMIs).


The hardware performance of the underlying sensors is therefore tied to the application’s perceived intelligence. Limitations in resolution, susceptibility to noise, thermal drift, or latency directly propagate into system behaviour, influencing stability, efficiency, and safety. In HMIs, for example, responsiveness and consistency are central to building trust and provide a vital interaction point with humans. Here, inputs must be detected and translated, ensuring that the system reacts in a way that aligns with user expectations while ergonomically fitting the end


application with HMI requirements in industry varying significantly from those in medical or consumer environments, for example. In autonomous systems, the sensing challenge shifts toward maintaining reliable perception under uncertainty, where environmental variability and sensor limitations must be managed without compromising system integrity. While the underlying AI models may even be similar across these domains, the sensing


Components in Electronics


requirements diverge


significantly. Systems that interact


directly with humans prioritize latency, smoothness, and repeatability, while autonomous systems place greater emphasis on precision, redundancy, and robustness. This divergence reinforces that sensing is not a uniform layer within AI architectures but a highly application-dependent function that shapes overall system performance.


How AI is reshaping sensor requirements


As AI systems evolve, they impose new requirements on sensing technologies. Integration, timing, and system-level reliability are becoming as critical as raw measurement capability, particularly in systems such as autonomous vehicles, industrial robotics, and smart industrial infrastructure.





One of the most visible shifts is the move toward digital outputs. In electrically complex environments such as automotive platforms, robotics, and large-scale industrial systems, digital interfaces provide improved noise immunity and enable deterministic communication with microcontrollers and processing units. This reduces ambiguity in signal interpretation and simplifies integration into centralized architectures, where multiple data streams must be synchronized and processed in real time.


Speed and diagnostics


Latency is another defining parameter. In closed-loop control systems, such as robotic actuation, delays in measurement and transmission directly affect stability and precision. In HMIs, latency directly affects responsiveness and the user’s perception of quality.


Built-in diagnostics and self-monitoring are


also gaining importance as systems become more autonomous. Sensors are increasingly expected to provide information about their own operating state, supporting higher-level decisions related to fault detection, redundancy management, and functional safety. In AI-driven systems, especially safety-critical ones, this extends beyond component reliability to system-level assurance, where the ability to detect and manage degradation is critical.


Redundancy and reliability are key Decreased human oversight is also driving the adoption of heterogeneous sensing architectures. A system cannot afford to lose visibility if a single sensing element fails, particularly in safety-critical environments such as autonomous driving or robotic automation. Instead, multiple sensing types can be combined to ensure continuity, allowing the system to maintain awareness even under fault conditions.


 environments


In parallel, increasingly electrified systems, including autonomous vehicles and industrial robotics, introduce additional challenges such as stray magnetic fields and electromagnetic interference. For AI-driven systems, where perception feeds directly into model inference and control decisions, corrupted measurements can propagate into incorrect outputs. Stray field immunity (SFI) therefore becomes essential to ensure that data remains stable and accurate, particularly in proximity to high-current conductors or switching power electronics commonly found in electrified vehicles and industrial systems.


www.cieonline.co.uk


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48