As humanoid robotic builders race to maneuver their machines from demonstrations and pilot tasks into factories and different working environments, one other problem is quickly coming into focus: find out how to manufacture these extremely complicated machines at scale.
Constructing a handful of working humanoids is one factor. Producing a whole lot or ultimately 1000’s of robots with constant high quality, predictable prices and dependable provide chains presents a really totally different engineering drawback.
Humanoids include dozens of actuators alongside motors, precision gears, bearings, sensors and machined parts, whereas lots of their designs are nonetheless evolving quickly as producers study from real-world deployments.
That creates an uncommon manufacturing problem wherein corporations might have to extend manufacturing whereas persevering with to change the product itself.
The problem is changing into more and more essential as funding in humanoid robotics grows and firms together with Boston Dynamics, Determine, Agility Robotics and a number of other main automotive producers put together for larger-scale deployments.
On this Q&A, Steve Ricketts, SVP of enterprise improvement at Misumi Americas, a serious provider of mechanical parts and industrial automation elements to producers and machine builders, and digital manufacturing and supply-chain firm Fictiv, discusses what occurs when humanoid builders make the transition from prototype to manufacturing, the place manufacturing bottlenecks are more likely to emerge, how prices may very well be introduced down, and whether or not the trade will ultimately converge round standardized parts and manufacturing processes.
Interview with Steve Ricketts
Robotics & Automation Information: Humanoid robotics corporations are demonstrating more and more succesful prototypes, however how totally different is the engineering problem of constructing one or 10 profitable robots from manufacturing a whole lot or 1000’s of them reliably and economically?
Steve Ricketts: It’s a essentially totally different drawback. A prototype proves the idea works. Manufacturing proves you possibly can repeat it. At one to 10 models, engineers can hand-fit elements, tweak tolerances, and work round quirks.
At a whole lot or 1000’s, each a type of workarounds turns into a supply of variation, price, and delay. Success depends upon design for manufacturability, tolerance stack-ups that maintain throughout batches, constant provider high quality, and meeting processes that don’t rely on a couple of expert individuals.
The groups that scale quickest deal with manufacturing as a design enter from the beginning, not a hurdle to clear after the design is completed.
R&AN: When robotics startups make the transition from prototype to manufacturing, the place do you mostly see issues emerge?
SR: Most frequently it’s designs that had been by no means created with quantity manufacturing in thoughts, and almost each different drawback traces again to that.
Elements that had been CNC-machined from strong inventory for a prototype might have to change into castings or molded parts, which adjustments tolerances, tooling lead instances, and value.
We additionally see tolerances specified tighter than the operate requires, meeting sequences which might be tough to repeat, and single-source parts that change into dangers at quantity.
High quality management is one other hole, as a result of inspection strategies that work for ten elements not often scale to a thousand. Catching these points early is way cheaper than redesigning later.
R&AN: Humanoid robots include unusually giant numbers of actuators, motors, gears, bearings, sensors and precision-machined parts. Are there explicit parts or manufacturing processes that you simply count on to change into bottlenecks as humanoid manufacturing volumes enhance?
SR: Actuators are the almost definitely strain level. Every humanoid wants dozens of them, and every actuator combines motors, precision gears, bearings, encoders, and housings.
Excessive-precision gearing, similar to harmonic drives and planetary reducers, and the machining behind them are capacity-constrained at this time. Uncommon-earth magnets for motors are one other supply-chain concern, as are specialised bearings and sensors.
On the method aspect, five-axis machining, precision grinding, and tight-tolerance inspection will pressure present capability. Count on a shift towards extra automated manufacturing and in-line metrology to maintain high quality constant as volumes climb.
R&AN: Conventional product improvement tends to maneuver from prototype to design freeze after which quantity manufacturing, however humanoid corporations are nonetheless altering their {hardware} quickly as they study from deployments. How do you manufacture at scale when the product itself might proceed altering from one manufacturing batch to the following?
SR: You want a producing method constructed for change. Which means utilizing versatile processes, similar to CNC machining, 3D printing, and comfortable or bridge tooling, for so long as the design remains to be evolving, and committing to arduous tooling just for steady subassemblies.
Modular architectures assist, as a result of you possibly can replace a forearm or a hand with out redesigning the entire robotic.
Robust revision management, clear communication with suppliers, and digital manufacturing platforms that may quote and produce up to date elements rapidly all maintain iteration from turning into chaos. The objective is to scale manufacturing in phases slightly than ready for an ideal design.
R&AN: There’s appreciable dialogue about ultimately bringing the price of humanoid robots right down to tens of 1000’s of {dollars}. From a producing perspective, the place are the largest alternatives to scale back prices?
SR: It is going to take a number of levers working collectively. Design simplification is the largest: fewer elements, fewer distinctive fasteners, and fewer tight tolerances the place they aren’t wanted.
Element standardization comes subsequent, since reusing actuator designs throughout joints concentrates quantity and lowers unit price.
Transferring from machined elements to castings, stampings, and injection-molded parts makes an enormous distinction as soon as volumes justify tooling. Quantity itself drives down provider pricing, and materials selections matter too.
Vertical integration might help with essentially the most essential subsystems, however most corporations will nonetheless depend on a robust provider ecosystem to get there.
R&AN: A few of the largest corporations getting into humanoid robotics have substantial manufacturing assets of their very own, whereas startups are rather more depending on exterior suppliers and manufacturing companions. Does that give established automotive and industrial corporations a big benefit because the sector strikes from prototypes towards mass manufacturing?
SR: Established producers do have actual benefits: provide chain leverage, manufacturing experience, capital, and expertise with high quality methods at scale.
However startups aren’t with out strengths. They transfer quicker, take design dangers that bigger organizations keep away from, and might construct new architectures with out legacy constraints.
Entry to manufacturing functionality can also be changing into much less of a barrier, as a result of on-demand manufacturing companions and digital platforms give smaller groups entry to capabilities that when required proudly owning a manufacturing facility.
I count on the winners to be whoever pairs nice {hardware} design with disciplined manufacturing execution, no matter firm measurement.
R&AN: If humanoid robots ultimately attain automotive-style manufacturing volumes, how totally different will the robots themselves have to change into from the machines we’re seeing at this time? Do you count on the trade finally to converge round extra standardized parts, architectures and manufacturing processes, or will humanoid {hardware} stay extremely proprietary?
SR: Right now’s robots will look fairly totally different from the amount variations. Count on fewer elements, extra built-in subassemblies, and designs optimized for casting, molding, and automatic meeting slightly than machining. I feel we’ll see partial convergence.
Commodity parts like motors, bearings, sensors, and fasteners will standardize, and actuators might comply with because the market matures, a lot because the auto trade standardized many parts.
Differentiation will doubtless keep proprietary in areas like software program, hand design, and total system structure. The trade will in all probability find yourself with a shared provide base supporting distinct merchandise.
