Robotics
Legged robots: what they are for
Quadrupeds became a commercial product before humanoids, and the applications are narrower and more sensible than the videos suggest.
Four-legged robots moved from research laboratory to commercial availability faster than most expected. What they are actually used for is worth knowing, because it is not what the marketing videos depict.
Why legs at all
Wheels are more efficient on flat ground by a wide margin. Legs are justified only where the ground is not flat.
The genuine advantages are stairs, rubble, ladders in some designs, thresholds, cable trays, and terrain with gaps a wheel would fall into.
In an industrial facility built for humans, those obstacles are common — which is the actual argument, and it is a modest one rather than a revolutionary one.
What made it work
Control. Dynamic balance on legs was solved through a combination of model predictive control and, more recently, policies trained in simulation and transferred to hardware.
Simulation-to-reality transfer has been the significant advance: training in massively parallel simulation with randomised physics produces controllers robust enough to work on real hardware, which was not reliably true a decade ago.
Actuators. Cheaper, higher-torque electric motors with integrated control replaced hydraulics, which were expensive, leaky and loud.
Sensing. Cheap depth cameras and lidar for terrain perception.
The actual applications
Routine inspection. The dominant commercial use.
An industrial site — a substation, a refinery, a construction site, a power plant — requires regular walking rounds: reading gauges, checking for leaks, thermal imaging of equipment, listening for abnormal sounds.
The task is repetitive, the route is fixed, and the value is in doing it more often than a human rota allows. Daily inspection catches problems that weekly inspection misses.
Note that the robot is not making decisions. It follows a taught route, collects data, and a person or an analysis system reviews it.
Hazardous environment entry. Confined spaces, areas with atmospheric hazards, post-incident assessment, decommissioning sites.
The value is straightforward: not sending a person.
Construction progress documentation. Walking a site on a schedule, capturing scans that are compared against the model to track progress and identify deviations.
Security patrol, with mixed results and real questions about whether it is better than fixed cameras.
Research, where the platforms have become a standard tool because they are available off the shelf.
What they are not doing
Carrying heavy loads over rough terrain, which was the original military motivation and was abandoned largely because the machines were too loud and too maintenance-intensive.
Manipulation of any complexity. Arms fitted to quadrupeds can open doors and press buttons, which is genuinely useful and a long way from general manipulation.
Operating unsupervised for long periods. Battery life is measured in around an hour of walking, requiring docking stations for autonomous operation.
The economics
Purchase cost is substantial, and the comparison is against a person's time for a repetitive task in an unpleasant environment.
The case is strongest where inspection frequency is currently constrained by labour availability, where the environment is hazardous, or where consistency of data collection matters — a robot takes the photograph from precisely the same position every time, which makes change detection far easier.
That last point is underrated and is a genuine advantage over a person with a clipboard.
Humanoids versus quadrupeds
The comparison is instructive.
Quadrupeds are stable, mechanically simpler, and do one thing — traverse terrain — well.
Humanoids attempt to be general, which requires balance on two legs and dexterous manipulation, both harder, in exchange for potentially using human tools and spaces.
The quadruped succeeded commercially first because it made a smaller promise and kept it. Whether the humanoid's larger promise is kept is the open question, and the quadruped's history suggests the answer arrives through narrow applications rather than general capability.
The public reaction
Worth acknowledging.
These machines provoke unease in a way wheeled robots do not, and deployments in policing contexts have met organised opposition in several cities.
Some of that is aesthetic and some reflects a legitimate concern about surveillance and the use of force. Manufacturers have responded with usage policies prohibiting weaponisation, which are commitments rather than controls.
