Insights: Case Studies

Catsuits, Reworking PolygonVR and Bleeding-Edge VR Tech – BTS with Neurogaming

Alternative Text

Dima Neiaglov | EU General Manager | Neurogaming

15 Feb 2019 | 4 min read

Who is Neurogaming?

Neurogaming was born from a joint venture between Wargaming, an award-winning publisher that is well known for working on World of Tanks, and VRTech, a group of companies offering a diverse portfolio of innovative virtual reality projects. Neurogaming was formed with the intention of creating a company that could become the ultimate business solution for location-based VR entertainment.

Founded in 2018, Neurogaming has already developed a number of cutting-edge VR experiences including PolygonVR, PlayVR and ArtVR. With the company’s rapid growth, Neurogaming have won a variety of accolades from within the industry. Most recently picking up an award for Rising VR Company of the Year at the second annual VR Awards in October 2018.

Here, Neurogaming’s EU General Manager, Dima Neiagalov, takes us through the making of PolygonVR and provides a peek into the state-of the art technology that made the project such a success with exclusive videos from behind the scenes. 

Stars and Constellations

To the professional community, Neurogaming’s PolygonVR might be curious as the first project to use Optitrack for full-body tracking and the first publicly available VR platform to use machine learning for skeleton solving.

Most free-roaming VR projects like The Void, Zero Latency or TrueVR Systems (see many more here) try to implement an archetypal mixture of the holodeck from Star Trek and Planet Doom from Ready Player One. Neurogaming was no exception—although if we honestly compared ourselves to competition, we admittedly produced more alcohol-induced videos in the process.

In 2016, full-body was practically invisible. VRcade and The Void, two household names of free-roaming VR at the time, didn’t have full-body capability. As many others, we started experimenting with ’constellations’—groups of optically recognizable markers. We used Vicon.

What seemed an example of stellar progress was quickly impeded by the most unfortunate realization: most people only have four limbs. And so, too many constellations were required to support a decent number of players, and too often they glitched because of ambient light or overlapping.

Optical Recognition, Inverse Kinematics and Catsuits 

The natural next step was to use optical recognition (aka ‘passive markers’) and inverse kinematics: an algorithm that can build your skeleton knowing the position of only some of your body parts. For the latter we used IKinema—and our friends at The Void continue to use this technology. This was the period of the first badass demos, our R&D director got to pose in a quasi-military outfit. You can check that out here, where he is painfully aware of being filmed:

Optitrack at the time was quite persistent against overlapping. If you saw the first enthusiastic reports from GDC or Gamescom, they likely came from this era. By current standards, tracking is too jiggly: 

Backstage pictures, of course, show a lot of sleepless faces of Neurogaming’s finest. The issue of optical tracking being too demanding towards ambient light and the process of having to tweak it so that every arena was practically light-isolated did not help matters. 

And so we had a tracking system, an inverse kinematics system and a working build of the game. Everything was almost ready to hit the commercial market, we only needed to spend a couple of months polishing and refining PolygonVR so it was ready for launch. The team were in a good place and ready to deliver the game on schedule, things were going exceedingly well. 

It’s only natural then, that 16 months later, when PolygonVR finally hit the market, the project had been almost completely reworked from the ground up. 

Utilising Machine Learning

IKinema makes brilliant products but they’re still someone else’s products and subsequently presented the classic case of ‘not invented here’ – unchangeable, and maybe a bit too costly. Mathematics within inverse kinematics is pretty intense and so the R&D team chose a different method of approach: machine learning.

Within the world of machine learning, a solver is trained by people simply walking around. It then learns that if such-and-such points of the body are like so, then the whole body must be like so—and applies this in VR. The outcome is a much more sensitive, accurate and bullet-proof skeleton solver.

The addition of active markers and a switch to StrikerVR guns propelled PolygonVR towards the mainstream business world, as did the addition of auxiliary software to manage PolygonVR as a business.

Of course, PolygonVR grew in other directions, too. Our arenas are now interconnected and broadcast-ready: we hope to become the leading supplier for VR eSports. We also hope to win customers by aggregating a lot of business data. But underneath it all, ultimately, is you. Our goal is to continue to make use of bleeding-edge technology to create the very best VR experiences we can. 

Here’s an exclusive sneak peek of some future stuff the team is working on: 

 

About the author:

Dima is EU General Manager at Neurogaming and VR Tech. He has been intimately involved with all of Neurogaming's projects since launch including PolygonVR, World of Tanks VR and ArtVR.