Autonomous vehicle testing in the simulator

LinkedIn +

On the surface, the rise of the autonomous car might appear to render human drivers a little superfluous in the development process. The reality of course is very different.

If you think about it, we’re at a fascinating historical junction. The whole history of the car, from the industrial revolution to the present has been driven by the desire to extend people’s mobility. That concept is deeply anchored in place – so much so that it’s now taken for granted. The extension of mobility is no longer the sole selling point to new buyers.

So how do you engineer a brand feel when the car is partially or fully autonomous? It’s safe to say this is something the car manufacturers are already looking into. Historically manufacturers have anchored their brand identity on the way their vehicles drive, so the advent of autonomous car will either displace that paradigm or morph it into something new. It may be that our perception of brand identity will be different in future, but the human being will still be at the center of things.

As more and more ADAS systems become a reality, it becomes harder to predict how they will corrupt or enhance the pleasure of driving the car. And the only way you can answer that is with a human being. Over the last few decades, vehicle design has migrated toward a heavier reliance on model-based development, but in a sense we’re almost seeing a return to subjective testing.

In many instances the best way to carry out this testing is with a Driver-in-the-Loop simulator and as control shifts towards autonomy the fidelity required from the simulator goes up quite a bit. Even things that we might classify as human factors research can actually be influenced quite heavily by the vehicle’s dynamics. This is particularly true when you’re dealing with experienced test drivers, who have a very finely honed sense of what to expect from a ‘real’ car.

Ride comfort is an interesting area, because it’s at the far end of the scale in terms of subjective fidelity. All the inputs to your senses are weighted differently compared to other experiments. For instance, our perception of comfort can be strongly influenced by the audio characteristics of the vehicle.

We’ve done a lot of work in this area. The subtleties of cavity resonance and wind noise can all be modeled and injected into the simulator to provide a level of fidelity that hasn’t even been possible until relatively recently. There’s also the tactile experience of feeling movement and vibration from the seat and the steering wheel.

On a more fundamental level, you need exceptionally precise control of the motion platform to relay small changes in the ride behavior. This is a notoriously difficult aspect to model accurately in a simulator, but with sophisticated motion control, pus low levels of friction and inertia our engineering class simulators can provide a very close correlation to real-world data.

Ultimately, simulator testing will never completely replace real-world observations – even if the cars themselves do become fully autonomous. Crucially though, they allow you to test much as possible in a laboratory environment rather than risking people and equipment in potentially dangerous situations. And as the automotive industry experiments with new ideas on ADAS systems and autonomous driving the need for authentic, subjective feedback will only increase.

6 June, 2016

Share this story:

Comments are closed.