This article was written and contributed to Engineering.com by Florian Haspinger, cofounder and managing director, Hololight. All opinions are his own.
Reviewing a 3D CAD model in XR can close the gap between an abstract 2D screen review and a costly 3D prototype. But in practice, the technology often disappoints as soon as it is exposed to real 3D data used in engineering. Models are too big and complex, and the polygon count far exceeds what a headset can handle on its own, so teams are forced to build simpler versions of their models to review them in XR at all. But that loses the very details needed to understand the model in space in the first place.
Often the first instinct is to blame the headset for not being powerful enough, ending the session early and leaving the impression that XR cannot deliver what an actual engineering review needs.
While the headset is the bottleneck, what has held XR back is that teams accepted this as a reason to compromise the model or dismiss XR entirely—when the rendering never had to happen on the headset at all.
The same compromise, every generation
Immersive engineering review is not as recent as one might think. Solutions appeared as early as 1992, with the CAVE, a room-sized cube with projectors directed at the walls, floor, and ceiling that was developed at the University of Illinois Chicago. Viewers wore LCD shutter glasses fitted with a position sensor, and the projections updated to match the tracked viewer’s perspective as they moved. The rendering was done by the system driving the projectors, not by the glasses. But since the compute stayed in this one room, the review did, too.
Early adoption was led by major companies in automotive and aerospace and largely stayed there, because the investment required put it out of reach for everyone else. Because the rooms were costly and needed a lot of space, accessibility was a real problem.
Data also had to be prepared before it could be reviewed. Getting a CAD model to display in XR was a largely manual process, and engineers spent time tessellating geometry, decimating meshes and removing hidden parts to meet the system’s requirements and limitations.
While XR headsets have since become a portable and far cheaper solution, they did not eliminate the need for data preparation, due to the limited on-board compute the small devices provide. What remains after tessellation, polygon decimation and other data preparation is not the model of record, but only a derivative.
This creates two problems. The first is timing. Data preparation takes considerable work, so the reviewable version often lags behind the original CAD data, which keeps being developed in parallel. Merging annotations and recorded problems back into the original data after the session gives even more versions to keep track of.
The second problem concerns decision confidence. Whenever a review session uncovers an issue, the reviewer can never be certain the problem exists in the original model or simply as an artifact of the simplification process. A review based on less than full certainty about model state does not produce a good decision.
XR stalled in engineering not on what a headset could render, but on everything teams had to take out of the model to get it there.
For a long time, teams have been forced to make compromises. Utilizing a projection room is costly and does not eliminate travel. Headsets were able to remove this problem, but still required model preparation, which made the review decisions less reliable. Plenty of teams eventually gave up and went back to the flat screen, where at least the complete model renders correctly.
Move the compute, not the model
The more useful move is to shift the focus away from optimizing around these limitations and instead approach the workflow differently. With XR streaming, the model and headset remain untouched. The application containing the model keeps running where it always has, rendering at full complexity, and only the pixels travel to the headset, while the original geometry never moves.
This changes what a review can be. An assembly is reviewable as soon as it’s modeled, in full visual quality, including materials and lighting, and without the need for data preparation or a second version to reconcile afterwards. A team can look at a revision the same day it’s made, allowing for faster iteration loops and less reliance on costly physical prototypes.
People are able to join a review session in XR remotely, drastically reducing the need for travel to align across teams and locations. The spatial context can also help outside stakeholders and customers unfamiliar with CAD software to grasp the design far more quickly than on a 2D screen, which makes for clearer communication and helps catch issues before they become late-stage design changes.
For enterprises looking to use XR streaming in their workflows, two things are worth considering. Streaming moves the requirement from the headset to the connection between the render machine and the device, so it is worth knowing what that connection needs before a review is scheduled rather than during it. The second is the render hardware itself, which has to be sufficient for the application, the models being reviewed, and the number of sessions running at once.
Not every workflow or review stage requires XR, but with streaming, XR becomes something you can reach for the moment a question needs it. Open a session, bring in whoever can answer it, wherever they happen to be sitting, and look at the actual model together with all its spatial context. The tool fits the review instead of the review being built around the tool.
The point of a design review is to make a decision, and that only works if everyone is looking at the model as it was actually designed. XR streaming is what makes that possible.
About the author
Florian Haspinger is co-founder and Managing Director of Hololight, the enterprise infrastructure layer for industrial XR. By decoupling XR software from hardware and streaming pixels rather than data, Hololight gives organizations a secure, scalable standard for deploying AR and VR at scale, trusted by partners including NVIDIA, Autodesk, and Microsoft, and customers such as BMW and BASF. Florian studied physics at the University of Innsbruck, specializing in aerospace engineering, and sees immersive technology as one of the most important and future-proof forces in global digitization.
The post Opinion: What’s actually holding XR back in engineering appeared first on Engineering.com.