HOLOFLAME: Dynamic orchestration, customization and delivery of Mixed Reality holographic media content at the network edge

HOLOFLAME focused on conducting experimentation and performance analysis over the Level7 infrastructure in Buseto Palizzolo. The focus was on the implementation of Augmented Reality (AR) scenarios using commercial off-the-shelf (COTS) Mixed Reality (MR) hardware. In particular, we successfully validated 3 design patterns (nearest playout, proxy cache playout and content placement) in two different scenarios for the cultural heritage and the agricultural domain: (a) a use-case for the visitors of the local agriculture museum that have dynamic access to AR content for the artefacts, and (b) a crowdsensing use-case that allows the users to upload and share their own content (images, video, etc.) and the trainers/tourist guides to create simplified AR scenarios (without coding at all, simply by uploading their content). Final tests took place in June.

Figure 1: (left) HOLOFLAME Test topology; (right) HOLOFLAME deployment

In particular, the project investigated the adaptation of an existing MR application of the company. The considered MR application retrieves and visualises AR/MR content about events including 3D maps, paths and routes, 360-degree videos, Points-of-Interest (POIs), etc. In the context of HOLOFLAME, the existing application has been extended with the capability of mobile crowdsensing by allowing users to enrich our database with new multimedia content that can be used by other participants and visualised in the MR application. By exploiting the unique service routing capabilities offered by FLAME, the users were able to build dynamic MR scenarios by uploading, customising and delivering AR content. The MR scenarios were decompiled in a series of service functions (for AR content, storage, execution, content delivery) that can be accessed via the network. A series of stress testing scenarios were applied to assess the performance of the FLAME platform and infrastructure under different requirements imposed by the MR application in terms of: (i) network capacity, (ii) network latency, and (iii) quality of user experience.

Figure 2: HOLOFLAME QR code stickers and AR-based content visualisation.

Various experiments were performed remotely in sandpit and at the replica Level7. We have validated and tested a methodology for remotely testing our scenarios over FLAME. We have successfully achieved the remote access to sandpit and Level7 through port forwarding.

Below, we analyze the steps of our methodology for sandpit. The same steps were applied for accessing Level7. We deployed our service functions and connected through ssh (using the domain names). In Figure , the red node ue20 can communicate with green nodes (e.g., web.holoflame-intellia.ict-flame.eu:5000) and by applying port forwarding you can access all the nodes of the network remotely (e.g., ssh -L 8000:localhost:5000 ubuntu@web.holoflame-intellia.ict-flame.eu -N -v).

Figure 3: Total time required with and without caching the AR content (statistics for image and 3D content)