Author: Merryn Ballantyne

Written by Dr. Katia Bourahmoune, Acting Co-Lead, Quality Assurance & Compliance program. 

In April 2026, a humanoid robot crossed the finish line of a Beijing half-marathon in fifty minutes and twenty-six seconds, faster than any human being has ever run that distance [1]. Months earlier, humanoid robots had performed incredibly complex, synchronized martial arts routines [2]. As the media coverage gained widespread attention, something more interesting than this engineering achievement emerged: a question, not yet fully formed, about what kind of world we are now entering, and whether we are entering it with our eyes open. 

The choice to build robots in the human form is sometimes caricatured as a vanity of engineers or a concession to popular culture and science-fiction media, however, its philosophical wager is of considerable depth. The world into which these machines are being released (its factories, hospitals, construction sites, and even homes) was designed for users that stand upright, use two hands, and react dynamically to the world around them. 

Traditional forms of industrial and collaborative robots were built for tasks in bounded environments, e.g. a wheeled platform optimised for a warehouse floor or an articulated arm for a single weld point on an assembly line. A humanoid robot carries an inherent optimism about general physical intelligence: the bet, or perhaps ambition, that a machine capable of inhabiting the full texture of human environments can in time respond to the full texture of human need. The ancient concept of Ziran in classical Chinese thought illuminates what the designers in this field are reaching toward. Ziran is often rendered as naturalness, or the disposition of things to accord with their own nature [3]. In the context of robotics, this can be found in the idea of building machines that fit the world as it is, rather than demanding the world be remade to fit the machine. 

Humanoid robots are now operating in production environments and shipping in volumes that would have seemed premature as recently as 2023. Venture capital investment in humanoid robotics exceeded three billion dollars in 2024, with reports of multi-billion market projections for the next decade [4]. What this momentum cannot easily tell us is whether the design assumptions underlying this transition have been adequately examined. The present dominant commercial logic treats humanoid forms primarily as a means of fitting machine labour into existing human infrastructure i.e. same floor plan, same tools, and minimal workflow redesign. That is a reasonable engineering position. It may also be eclipsing, earlier than is wise, questions around whether the humanoid is best understood as a substitute for human presence or as a platform for augmenting it. 

It is worth noting that the world these machines are being designed to inhabit was itself built around the human body. Every dimension of that infrastructure, accumulated across two centuries of industrial development, was calibrated to the physical limits and capabilities of the biological human form. While previous waves of automation reshaped work around the machine, the humanoid, at least in aspiration, inverts that relationship. In doing so, it raises a concern that the industrial revolution never had occasion to face: What becomes of the human body’s centrality to working life when the physical form that justified building the world around it can be replicated, scaled, and indefinitely reproduced? That this question is now being asked simultaneously in boardrooms, parliaments, and papal encyclicals is perhaps the clearest measure of its weight [5].  

What the field of collaborative robotics has understood for some time (and what the humanoid moment is now forcing into general visibility) is that matching human physical capability, however necessary, is not sufficient. Harder questions concern the relationship between human and robot: what kind of human-robot partnership produces durable, humane, and useful outcomes, and under what conditions workers can reasonably extend trust to machines working beside them. Those questions shaped decades of research into human-robot interaction and collaboration and the work the Australian Cobotics Centre has been part of since 2021. How much humanoids come to define the next chapter of that work is ultimately a question research and humanity will have to answer. 

 

Call for Participation

The Australian Cobotics Centre is calling for experts across academia, industry, and government to participate in a research study at the University of Technology Sydney aimed at developing a clearer definition of collaborative robots. Participation involves an online discussion followed by a brief activity to rate statements about cobots. Your input will directly inform how the field defines and frames human–robot collaboration. 

More information and EOI here:  EOI and Consent Form (https://forms.office.com/r/YSYQPWqD8X) 

 

References and Further Reading:  

[1] Harmon, K. (2026). A humanoid robot beat the human half-marathon record at a Beijing race. But what did it actually prove? Scientific American.  

[2] Unitree Robotics. (2026). Kung fu meets spring: Unitree Spring Festival Gala robots present “Cyber Real Kung Fu” in the year of the horse [Press release]. PR Newswire.  

[3] Cleary, T. (Trans.). (1992). The essential Tao: An initiation into the heart of Taoism through the authentic Tao Te Ching and the inner teachings of Chuang-tzu. HarperCollins.  

[4] Goldman Sachs. (2024). Humanoid robots: A $38 billion market by 2035. Goldman Sachs Research.  

[5] Leo XIV. (2026). Magnifica Humanitas [Encyclical letter]. Dicastery for Communication, Holy See.  

 

 

Industrial automation has traditionally been built around a fundamental limitation: robots cannot “feel”. Sensing physical interaction with a workpiece or environment has historically required expensive hardware, making such capabilities impractical for many industrial systems. As a result, conventional industrial robots typically operate in isolation, executing preprogrammed motions without direct awareness of the forces they encounter. 

The development of collaborative robots (cobots) introduced the ability to sense internal forces and detect collisions, allowing for safer human-robot interaction. However, true physical awareness requires external sensors. When equipped with exteroceptive sensors, such as those that measure forces or vibrations, robots can respond to external conditions like a changing workpiece. This capability expands robotic automation into applications that require both force sensitivity and precision, including complex finishing operations such as grinding and polishing. 

Grinding remains one of the most physically demanding tasks in metal fabrication. The process requires a balance between force and precision; too little pressure slows production, while too much risks damaging the part or wearing down the tool prematurely. These characteristics make grinding a promising candidate for automation. Many manufacturers pursue robotic grinding not only to address rising labour costs and workforce shortages, but also to achieve process consistency and repeatability. 

However, implementing robotic grinding typically requires high-end sensing hardware. These systems often rely on force/torque sensors to measure the interaction between the tool and the workpiece, enabling robots to maintain the controlled force necessary for precision finishing. These sensors can cost upwards of $4,400 USD. For many small and medium-sized enterprises (SMEs), particularly in Australia, this cost represents a significant barrier to entry, turning automation into a financial hurdle rather than a competitive advantage. 

Recent research by PhD candidate, Zongyuan Zhang and his supervisory team suggests that robots may not need expensive sensors to achieve force awareness. Human operators performing grinding tasks often rely on subtle auditory cues, like the pitch and vibration of the tool, to judge the quality of contact with the material. Experienced machinists can detect changes in force or tool wear simply by listening to the sound of the process. Inspired by this intuition, researchers have begun exploring whether similar information can be extracted using low-cost acoustic sensing combined with machine learning. 

The proposed Acoustic Feedback Robotic Grinding (AFRG) system (see Figure 1) demonstrates how this approach can work in practice. Instead of measuring force directly, the system monitors the acoustic signature of the grinding process. A single contact microphone is mounted to the tool bracket, capturing vibrations transmitted through the structure of the tool while filtering out much of the ambient noise present on a factory floor. 

The captured signal is processed by a specialised neural network known as PSDRegNet, a two-dimensional convolutional neural network designed to estimate the grinding force. By learning the complex relationship between acoustic patterns and grinding forces, the model can estimate the interaction force in real time. This data can then be used to adjust the robot’s behaviour online. Since the system learns this relationship directly from data, it avoids the need for rigid mathematical models that would typically govern robotic finishing processes. This flexibility allows the same system to adapt more easily to different materials, tools, and process conditions, reducing the time and engineering effort required to reconfigure robotic cells for new tasks. 

Another challenge in robotic finishing is tool degradation. As grinding discs wear down or become clogged with material, their cutting efficiency declines. Robots that rely on fixed paths or constant force setpoints often struggle to compensate for this change, leading to inconsistent material removal over time. In experimental trials conducted on hardened stainless steel, a material known for accelerating tool wear, the AFRG system demonstrated a fourfold improvement in grinding consistency compared to traditional force-based control. Since the acoustic model captures tangential force information closely related to the material removal rate, the system can maintain a stable finishing process even as the physical properties of the grinding disc change. 

The implications extend beyond grinding. If meaningful process information can be extracted from inexpensive sensors such as microphones, accelerometers, or cameras, machine learning may enable a new generation of low-cost perceptual capabilities for industrial robots. Instead of relying on specialised hardware for every sensing task, robots could infer key physical variables from readily available signals. 

For the Australian Cobotics Centre, this approach demonstrates a cost-effective pathway for quickly upgrading existing industrial infrastructure, something important for many Australian SMEs. Many legacy robots are position-controlled, following a set of predefined positions without sensing the forces involved in the task. Retrofitting these systems with force sensors can be costly, but in contrast, an acoustic sensing system can be integrated with minimal modifications, offering closed-loop force control at a fraction of the cost. 

More broadly, this work challenges the assumption that precision automation must rely on expensive hardware. By combining off-the-shelf sensors with machine learning, it becomes possible to convert robots from pre-programmed machines into adaptive systems capable of responding to their environment. 

 

 

 

 

 

 

 

Acoustic Feedback for Closed-Loop Force Control in Robotic Grinding, Zongyuan Zhang*, Christopher Lehnert, Will Browne, Jonathan Roberts