Defining fidelity in Simulators
A major concern in planning a simulator for training is whether to go all-out and implement custom hardware, making the simulator as close a copy of the machinery being simulated as possible, or to create a lightweight representation that has functional equivalents for major systems, saving on cost what you lose in fidelity. Three parameters define the overall fidelity and accordingly, cost of development, in simulators.
1. Functional fidelity is a measure of how well we emulate the user input -> System Response -> Output to User chain. Creating functional equivalence can be as simple as coding a few equations to approximate physics, or as complicated as replicating the entire code and hardware responses of a complex control system. This is also the area where most of the unavoidable cost and development time arise.
2. Audio-visual fidelity is a measure of how realistically the environment is presented to the user in terms of visuals and audio cues. Budgeting for Visuals and Audio is a well-trodden road, and there are multiple methods to reduce costs and to cater to your particular budget.
Example of Low vs High Visual Fidelity in a Flight Simulator
3. Physical fidelity is a measure of how closely the interface to the simulator resembles that of the actual simulated machine. E.g. when simulating a switch, a touch interface has a higher physical fidelity than one based on mouse movements, while an actual physical switch has the highest fidelity. Perfectionism in this aspect of simulator development can greatly inflate your budget and you can save a lot of cost and reduce risk considerably if you implement the interface smartly.
Simulating a spacewalk on the ISS, low vs high physical fidelity
Cost breakdown based on simulation quality levels
So how does a simulator project breakdown look like based on the three parameters above? Based on my own experience and reports from friends who work on similar projects in a wide variety of fields, this is what I have found:
Simulation project cost estimation based on how much fidelity is required
Note the above are just rough estimates of the cost breakdown, but there are a few trends that are easy to spot:
1. Increasing a project’s fidelity, particularly physical and visual fidelity by a step can increase costs by a lot.
2. Physical fidelity costs rise very quickly the closer you try to emulate the original object
3. Audio/visual fidelity costs go up fast too, but not as badly as physical fidelity.
This means that in order to keep costs low, you need to limit physical, and audio/visual fidelity to what is absolutely necessary. Additionally, if you can achieve our training goal with a lower overall fidelity, you should not aim higher initially. I would suggest developing a low-fidelity simulator, and getting real-world data on training efficacy, before moving up to the next step. While some of the visual or physical assets will not be reusable for a later upgrade, most of the functional simulation will.
Now that we have established increased physical fidelity as the major culprit in inflated simulator costs, let’s see what can be done about it.
How much physical fidelity is enough?
The answer to this question really depends on the situation. Mimicking the hardware as much as possible allows trainees to gain muscle memory and test the limits of their physical abilities in a realistic environment. For critical missions, like those performed by astronauts in orbit, recreating as much of that environment as physically possible makes sense. On the other hand, a race for maximal hardware parity can lead to budget over-estimations and hence project abandonment. The efficacy of simulators as a learning tool has been proven by many studies, thus it is worth looking at possibilities to capture the essence of the learning experience they provide, at the lowest possible cost to the organization. To understand what features can be scrapped without losing a whole lot of training value while seriously cutting down cost, let’s have a look at the benefits of high hardware fidelity:
- The tactile sense of realistic hardware controls helps train muscle memory. (commercial haptic feedback devices are still way behind the curve compared to other user interface technologies)
- Touch – there is no way to convey low-frequency vibration. (high frequency vibration can be conveyed by sound), etc. (e.g. in an HVAC or car repair simulator)
- Secondary and tertiary effects that would be too hard to simulate in software. E.g. a very cold pipe in the real world will form condensates on it – this would have probably not been simulated in a normal simulator, thus the cue would not be available to the trainee.
- Intuition is developed far better due to the multi-sensory environmental cues in a hardware based simulator and in the real world, compared to software sims.
- Software sims depend on a level of abstraction and not all actions involved will be simulated, e.g. in a 2D simulation of a rotary gauge, the parallax effect will be “abstracted out” hence the user never learns how avoid this effect. (3D sims can of course teach this)
- It is doubtful that cheap generic hardware for simulating full-body vibration, angle of body, gravity, acceleration, etc. will become commercially available. This means that ultra-realistic flight simulators e.g. where such effects are important (e.g. an airplane shaking violently might require input from the pilot) will still require expensive hardware.
Aside from cost reduction, moving towards software-only solutions with a lower physical fidelity offers other benefits:
- Software-only controls allow for easy modifications without incurring any manufacturing of custom parts
- software-only solutions can be installed on multiple machines and used by many users concurrently
- software-only simulators ride the wave of advancement driven by gaming and software technology – much faster than advances in manufacturing, miniaturization and machining
- software-only solutions allow for simulating machinery that is still in early design and planning stages and provide invaluable information to interface designers
- software based solutions generally have a much lower cost to operate and a negligible maintenance cost
If you are looking to create a simulator for a vehicle, equipment or a skill/process, start by implementing a low fidelity (particularly physical and visual) solution. In most cases this can be done on a very low budget in the tens of thousands to hundreds of thousands of dollars. Once you have evaluated the training efficacy for your target audience, it is easy to ramp up visual and in particular, functional efficiency. Avoid increasing physical fidelity until you have reasonable proof that it is absolutely critical.