The competing Virtual Reality ecostructures currently in intense development – such as Oculus Rift, Samsung Gear, HTC Vive and Microsoft HoloLens – are all addressing the crucial issue of spatial positioning in diverse and innovative ways – and for a multiplicity of reasons, ranging from the pragmatic to the cynical.
Operating in proprietary (and often prototype) environments, and servicing different visions of VR’s future, tethered and untethered, VR manufacturers have two prime motives towards making sure that the virtual worlds they enable line up with the real world with zero latency: the efficacy of their own systems and the potential to develop dominant standards which are either potentially profitable (in terms of licensing) or prestigious (in the event of open sourcing).
Establishing where a VR device is positioned in three-dimensional space at any time can be addressed in a number of ways, or a number of combinations of these methods. The simplest incarnation of VR positioning uses entirely local coordinates, in the form of the gyroscopic and accelerometer readings inside smartphones. These, combined with GPS coordinates also incorporated into the phones, make external VR experiences such as Pokémon Go possible, and also facilitate taster systems such as Google Cardboard, which are not expected to involve more than small geographical movements.
By contrast, Augmented Reality (AR) systems such as Microsoft’s Hololens are aimed at far more intense and specific interactive virtual (or semi-virtual) environments, where movement outside of the target arena is not anticipated, and where the experience must be prepared with specialised space-scanning equipment to account for occlusion (furniture and other obstructions which must be dealt with when situated ‘behind’ the simulated elements). If the promotional videos are to be believed, the Hololens system in particular is showing great interest in RGBD-based depth-mapped evaluation of environments – basically attempting to map environments just by looking at them.
AR and VR also make extensive use of infrared cameras, working in concert with sensors in the proprietary headsets, to establish spatial positioning – though this obviously adds additional limitations to the scope of the environment and the cost of developing it.
But none of the solutions currently in development promise anything like a ‘universal’ standard for calculating pitch and yaw in real-time; the Hololens model cannot easily exit its studied target environment, and the Pokémon Go model cannot provide the kind of occlusion capabilities that promise genuinely interactive VR and AR scenarios.
Towards remedying this, a researcher from Stanford University has proposed a system which uses commonly-available WiFi signals to provide a kind of ‘satnav for virtual worlds’ – a framework which could potentially be adopted by a number of different commercial systems, and which could greatly increase the range of genuinely interactive environments.
Manikanta Kotaru’s WiCapture system addresses a number of limitations with visually-based mapping such as that used by the HoloLens, since this method relies on visual texture and often fails when faced with a featureless mass such as a white wall. WiCapture utilises WiFi packets from standard WiFi access points via special algorithms which mine the Channel State Information (CSI) in the packets to extract the necessary positional information which can be fed back into headsets or other apparatus.
The author describes WiCapture as a ‘commodity WiFi-based position tracking system would potentially enable VR on mobile devices and revolutionize number of applications that can be enabled on top of motion tracking.’
Significantly the system is resistant to occlusion and requires little more than a WiFi chip in the receiving equipment – a significant factor, since a great deal of the complexity of tethered equipment such as the Rift is taken up with calculating this problem by other means.
WiCapture registers the change of position by calculating the phase change of the CSI between packets, and includes novel methods of compensating for the old problem of lack of time synchronisation between facets of WiFi infrastructure.
The system was created with commodity Intel 5300 WiFi chips, and results in precision levels of 0.25mm, and tracking error levels not exceeding 0.88cm.
WiCapture currently has higher latency than existing commercial systems, since the spatial calculations are taking place over a network, but the intention is to address this in future research by using IMU sensors to correct drifts.