Augmented reality
Augmented reality keeps the physical surroundings in view and adds digital objects or information aligned in three dimensions with what the person is facing, responding as they move. Ronald Azuma defined it in 1997 by three characteristics (it combines the physical with the virtual, it is interactive in real time and it is registered in 3D) without tying it to any type of display.
Fyware's XR glossary, with the source behind each definition.
Augmented reality keeps the physical surroundings in view and adds digital objects or information aligned in three dimensions with what the person is facing, responding as they move. Ronald Azuma defined it in 1997 by three characteristics (it combines the physical with the virtual, it is interactive in real time and it is registered in 3D) without tying it to any type of display.
Registered means aligned: the virtual object holds its place in space while the person walks around it. With that condition Azuma left two cases out. A movie with digital effects is not interactive, and a flat label laid over a camera image is not combined with the surroundings in three dimensions. Some researchers of the time required a head-mounted display, and Azuma chose not to, so his definition admits anything from a monitor to a see-through headset.
Against virtual reality, the difference is what the person keeps: the surroundings stay in view and the technology adds to them. Against mixed reality the border is blurrier. In commercial use, mixed reality is the augmented reality whose content recognizes surfaces and objects and hides behind them, and this glossary uses that sense.
The technical problem underneath is registration, keeping the virtual where it belongs. Azuma counted it among the most basic problems limiting the field and put numbers on it. With 100 milliseconds between the moment the head is measured and the moment the matching image appears, and a head turning at 50 degrees per second, the virtual object lands 5 degrees off: almost 60 millimeters (about 2.4 inches) at an arm’s length of 68 centimeters, enough to send a technician to the wrong bolt. He named system delay as the largest single source of registration error, ahead of all the others combined. One way out he describes uses markers, such as LEDs or printed patterns, whose positions the system knows in advance; while at least one stays in view, a video system can align to within one pixel. Markerless tracking recognizes the floor or the shape of the object instead. Twyn, which Fyware distributes in Mexico, identifies the part from its own geometry against the CAD model, through the model tracking of its maker, Visometry.
The device follows from what the hands are doing. On a phone or tablet the person holds the screen, which works for inspecting a part or showing a machine on a buyer’s floor. In maintenance, with both hands on the equipment, the instruction works better on a mixed reality headset that leaves the machine in sight. Azuma already listed the maintenance and repair of complex machinery among the applications in 1997, with 3D drawings over the equipment that show the task step by step, and described a Boeing prototype that guided a technician building the wiring harness of an airplane’s electrical system. Before asking for a quote, it pays to set how much error the task tolerates: deciding whether a line fits in a building allows more drift than comparing a weld with its CAD, and that tolerance decides whether a phone is enough or model-based tracking is needed.
Target Robotics, a robotic production line integrator based in San Luis Potosí, sells its machinery with an augmented reality model that the salesperson places at full scale on the buyer’s own floor from a tablet. The salesperson points the tablet at the empty floor and, once the placement reticle appears, the complete line stands there with its conveyor and its arms, so the buyer can walk around it and see how much of the plant it takes. The model comes from the engineering design the line is built from, prepared to run smoothly on a handheld device, and Fyware also delivered a WebXR version that opens in a browser with nothing to install. Target Robotics reports one fewer meeting in the sales cycle, the one that used to go into explaining how the line fits in the customer’s space.
«1) Combines real and virtual 2) Interactive in real time 3) Registered in 3-D»
Azuma defines augmented reality by three characteristics and leaves it open to different kinds of display.
«lag of 100 ms [...] the angular dynamic error is 5 degrees.»
With 100 milliseconds of lag and the head turning at 50 degrees per second, the angular error reaches 5 degrees, almost 60 millimeters at an arm’s length of 68 centimeters.
«3-D drawings superimposed upon the actual equipment, showing step-by-step the tasks»
Azuma already counts the maintenance and repair of complex machinery among the applications, with 3D instructions over the equipment and a Boeing prototype for building aircraft wiring harnesses.
«an otherwise real environment is "augmented" by means of virtual (computer graphic) objects»
Milgram and Kishino place augmented reality inside the continuum: a physical environment to which computer-generated objects are added.
Ronald Azuma set them in 1997: it combines the physical with the virtual, it is interactive in real time and it is registered in 3D, which means aligned with the space. A movie with digital effects fails the second test, and a flat label over a camera image fails the third. The definition does not require a headset, so it covers anything from a monitor to see-through glasses.
It puts the instruction on the equipment, step by step, at the point where the work happens; Azuma described this in 1997 with 3D drawings over the machine that can also be animated. When the technician needs both hands, a mixed reality headset fits better. When the job is checking a part against its CAD, a tablet with model-based tracking, such as Twyn, works right at the line, and in sales a full-scale model on the buyer’s floor answers whether a machine fits.
It anchors content without a printed code or image: the device recognizes the floor or the shape of an object and fixes the model there. That is how Target Robotics places its production line on a buyer’s floor from a tablet, and how Twyn recognizes a part against its CAD model. With markers, Azuma noted, video systems can align to within one pixel as long as at least one marker stays visible.
Tell us what you want to solve in your plant. Fyware designs extended reality around the equipment and constraints of your operation, and builds it to grow from a pilot to the whole site.
Tom Klukowski, Fyware Partner · Valcre Marketing We reply in under 24 business hours, with scope, timeline and investment range.In the meeting you see the simulator running and what it would take to bring it to your plant, with scenarios adapted to your layout. And if you need to sell it internally, this page is your presentation: share it with your team.
