Keeping simulator sickness down

“Sickness induced by Virtual Reality (VR) devices poses a genuine threat to the viability of this new technology and its potential products.”

This was written by Eugenia Kolasinski for her doctoral thesis at the University of Central Florida in 1996. Even earlier, in 1992, Frank Biocca had asked “Will Simulation Sickness Slow Down the Diffusion of Virtual Environment Technology?” writing that “The widespread diffusion of immersive virtual environments (VE) is threatened by persistent reports that some users experience simulation sickness, a form of motion sickness that accompanies extended use of the medium.’

Now, more than 20 years later, with the leaps forward in the development of immersive technology, are we any better off? Perhaps in the 1990s when virtual reality (VR) was nowhere near as advanced as it is today, people were almost bound to get sick. Surely, today now that everything is faster, smoother and slicker, simulation sickness wouldn’t be a problem?

Simulation sickness is not a problem confined to the hardware and software of any particular manufacturer. It occurs because present day technology inevitably involves a change in the relationship between what we perceive (specifically see) in VR and what we feel in our bodies while perceiving. In physical reality there is a tight coupling between the movements of our body (especially our head and eyes) and the perceptual consequences. If you turn your head your vision is ‘updated’, for all intents and purposes, exactly in sync with your physical movements. Everything works together – your muscle and joint movements, your vestibular signals (your sense of balance and spatial orientation that keeps you upright including responses to acceleration and deceleration) and your whole proprioceptive system, which signals the movements and disposition of the body. Let’s add to that the visual system which maintains a relationship between accommodation (the ciliary muscles change the lens shape allowing you to look at near and far objects, pupillary muscles, that control the dilation/contraction of the pupil) and vergence (how the eyes rotate inwards or outwards as you look at near or far objects). In virtual reality, the tight coupling between all of these systems is disrupted to a greater or lesser extent.

Vergence is fixed with respect to the display when you’re wearing a head mounted display (HMD), but accommodation may still function and so the relationship between the two is broken. So, when you move your head, depending on the latency of the head-tracking system, the displayed images in VR won’t be updated at the same rate as when you look around the real world. Moreover, if the rendering system has a low frame rate (the speed at which each new image is displayed), you will see slower changes in the display than would expected caused by your movements. On top of this, the actual computed image updates may vary over time – for example, imagine you are in a bare virtual room and then ‘step out’ to the virtual street where there are thousands of virtual people walking around. This may result in the need to render millions of polygons per frame. If the graphics card can’t keep up with this, then there will be a further degradation of image display rate.

As I have written about before, in VR presence (the illusion of being in the virtual place), depends on the extent to which natural sensorimotor contingencies – how sensory activity depends on the activity of the perceiver – are supported by the VR system. But if the realisation of sensorimotor contingencies is as bad as it’s been suggested above, how can anyone use VR, let alone experience the illusion of presence in a virtual environment? The critical word is ‘extent’ – most of these problems can be alleviated. For example, if you stay in the same position and translate or rotate your head, you probably won’t experience sickness provided that the head tracking latency is low and the image refresh rate is high enough.

Most head tracking systems work at around 60Hz. Frame rates are at least that and often much higher, although there is still the problem that complex scenes may suffer because the graphics card can’t keep up, and they can’t render at the image frame rate. This is very much application dependent. I have been working in VR continuously since the early 1990s, and in the vast majority of our studies we have avoided inducing simulator sickness. This is primarily because we rarely move people through virtual space.

Moving in virtual space

A fundamental problem occurs when you move or are ‘translated’ through space without the actual corresponding body movements. Your eyes tell you that you’re moving through space, but there are none of the accompanying signals that should go with it. You don’t have any feelings of acceleration or deceleration and no corresponding vestibular signals (from your balance and spatial orientation). If we really have to move someone in VR to a different location in space, then we ‘teleport’ them rather than attempting to move them continuously. About three years ago I almost abandoned one study because it was necessary for people to move through the environment and in just a few seconds of the experience I became quite sick. We overcame that problem by very substantially slowing down the translation through space, and especially rotations [1] . After making those changes, only one of the 16 people involved abandoned the experience because of the feeling of sickness.

How common is simulator sickness?

The incidences of simulator sickness depend on the application. If there are no forced viewpoint translations, and there is six-degree of freedom (6 DOF) head tracking with low latency, then the incidence can be zero. If there are rapid viewpoint translations through the environment then the incidence may be near 100%. A 1995 survey by Kolasinski reported that about 60% of participants experienced some form of simulator sickness which was highest after 20 minutes of exposure. A 2008 study by Sara Sharples and colleagues found between 60-70% of participants experienced simulator sickness symptoms with a HMD. In a recent (2017) study which used the Oculus Rift DK2, Ryan Tyrrell and colleagues found that amongst 20 people in a control condition (non-patients), 60% had no simulator sickness symptoms, and 30% had mild or moderate symptoms. Fifty percent experienced these symptoms at mild or moderate intensity. However, assuming a high quality, low latency, high framerate VR system, the level of simulator sickness depends very much on the application (and also there are important inter-personal differences).

The most widespread instrument for measuring simulator sickness is the ‘Simulator Sickness Questionnaire’ (SSQ) by Robert S. Kennedy and colleagues, a self-reporting tool. Another method is to test postural stability before and after VR exposure, for example the work by Hironori Akiduki and colleagues in 2003. The participant stands with eyes closed on a board (e.g., a Wii board) and the amount of spontaneous sway is measured. If this is substantially different after the VR experience compared to before it indicates postural instability. This is an ‘objective’ measure compared to the SSQ self-report, but certainly both methods offer valuable information.

How can it be prevented?

Various techniques have been tried to alleviate simulator sickness. The most obvious is to simply never translate the participant’s viewpoint through the scene unless they are also actually physically moving. If it is absolutely essential to do this, then move them really slowly, especially avoiding changes in acceleration and rotations as far as possible. If it is simply a matter of getting someone from A to B, and the view in-between is not important, then simply ‘teleport’ them from point to point.

Six-degrees of freedom helps alleviate feelings of sickness in VR because with three-degrees of freedom you only get correct updates for head rotations (turning your head left or right). If you bend down, or move your head forward or to the side (head translation) nothing changes in the display, increasing the chances of you becoming sick. This is also a fundamental difference between model-based VR (based on computer graphics) and 360 degree VR based on video. In the latter your head is typically at one spot only and only head rotations matter. Bearing in mind that doing a pure head rotation without any head translation is almost impossible, this is why you might feel more queasy in some 360 degree VR than model based (depending on the application).

Using a treadmill so participants actually carry out the actions of walking may also help. E.H.Sinitski and colleagues (2018) reported that ‘25% of participants still experienced slight symptoms of simulator sickness after immersion, including eyestrain, headache, difficulty focusing, and dizziness. Although SSQ scores were greater after immersion, all participants were able to successfully complete the study without exceeding ‘slight’ or acceptable levels of simulator sickness symptoms.’ This study is encouraging, if not ideal. Using a treadmill (or indeed simulating one through walking-in-place) may reduce simulator sickness, since the visual flow is matched with an action similar to walking. However, it cannot deliver proprioceptive and vestibular feedback associated with actual locomotion, and of course, designing an application that only works with a treadmill may not be practical in many cases.

An increasingly popular method is to try to produce this missing vestibular feedback artificially. Electrical current is applied to the mastoid processes (part of the bones roughly behind the ears) to stimulate the appropriate nerves in sync with the movement through space (perhaps not one to try at home!). This technique tries to link together the visual and vestibular responses, which are broken in VR, and has shown some success in a driving simulator application, and recently has been refined and extended (for a projection display).

In 2016, Ajoy Fernandes and Steven Feiner reported that simulator sickness can be reduced by dynamically changing the display field-of-view, for example, narrowing the field of view when the participant is moving quickly through an environment, and making it wider when they are stationary. They argue that this manipulation does not reduce the reported sense of presence. These are interesting results and it is a technique that can be relatively easily implemented. However, although simulator sickness was statistically lower compared to a control group that did not have this manipulation, the difference was not that great. Most importantly, the questionnaire that was used to assess ‘presence’ does not directly measure it all, but only possible side-effects of presence.

In the work conducted by Eugenia Kolasinski that I discussed earlier, a number of predictors of simulator sickness were suggested. She found that 35% of her 40 participants experienced some level of sickness, and that the propensity of individuals to experience simulator sickness could be statistically modelled by four variables (sex, age, pre-exposure postural stability, and ability at mental rotation tasks) with a relatively high degree of predictive power. The specific results are not as important as knowing that individual differences matter and the likelihood of experiencing simulator sickness in response to the same stimuli may differ substantially between individuals. The more we understand this the more we can give warnings and advice to those with the greatest risk.

Virtual reality is at another crossroads. The first was in the 1990s when in spite of its popularity as a concept, few people could experience it because of the cost. Today there are various other factors that will determine the success or otherwise of VR – not least being the availability of novel content that exploits the power of VR as a medium in its own right. Overcoming simulator sickness will be another vital determinant of the success of VR.

My advice to VR participants: if you feel the slightest hint of simulator sickness whilst using virtual reality, stop immediately. Close your eyes and take off the head mounted display. Once simulator sickness takes hold it can continue for hours afterwards, and only a night’s sleep is sufficient to entirely remove the symptoms.

As for VR developers, my advice would be not to produce scenarios that are bound to produce sickness in most people; those that move, twist, turn, and accelerate or decelerate the viewpoint rapidly through virtual space. It is self-defeating to put people into a VR environment, especially for the first time, like a simulation of a roller coaster or flying through space, as it is bound to make them sick, and they never want to do it again.

VR is an amazing medium for enhancing the lives of people. It doesn’t need to be at the cost of making them sick.

 

[1] Max linear speed forward = 3m/s; Max linear speed backwards = 1m/s; Forward acceleration = 0.1m/s2; Backwards acceleration = 0.1m/s2; Max angular speed = 5º/s; Angular acceleration = 5º/s2.