AR with Edge Computing: Visualisation of International Space Station (ISS)
A common request for many industrial AR adopters: How to visualize data-intense 3D objects fast and in high quality? This is a challenge, which needs to be met whether it is for design review, prototyping, visual support during complicated work processes, or simply for optimized communication and collaboration purposes.
If we look at one of the most popular devices for industrial use, Microsoft HoloLens, it does not take long to see that the benefit of such a self-sufficient head-mounted display (HMD) also comes with the disadvantage of limited memory and processing power. Other MR/AR devices are also not immune to this problem.
In order to fit within the memory capacity of these HMDs and allow applications to run smoothly, CAD models often need to be prepared with polygon reducers. This in turn, could cut down the quality of the 3D model to such an extent that the AR experience is not useful anymore.
Industrial players are increasingly demanding applications of greater complexity which require a higher level of performance in general. To resolve these competing issues, Holo-Light decided to explore the possibilities of remote rendering technology via edge computing.
Outsourcing Computing Power
The basic idea behind it lies in the outsourcing of demanding work processes like calculations, application logic, or content rendering. This means that the computing power does not need to come from the HMD itself, but instead can come from an external server which can more efficiently complete the processing.
Deutsche Telekom supports Holo-Light in Prototyping program
Deutsche Telekom is working on building a complete infrastructure of edge servers to support the realization of edge computing. Their Low Latency Prototyping program, organized by Telekom’s incubator hub:raum, provided a platform which allowed us to make productive advances in this field. It also has made it possible for Holo-Light to create a simulation of the solar system and a realistic, scale model of the International Space Station (ISS) rendered in real time.
This edge computing prototype could process 14 million vertices, 80 million polygons, and 4.5 GB of random access memory (RAM) at a rate of 40-60 frames per second, while a self-sufficient HoloLens can only run smoothly below approx. 200,000 polygons.
This means that Holo-Light is able to visualize models at least 50 times greater in size with remote rendering via edge computing than with the Holo-Lens alone.
