Article 2 of 7 in our Content series: Dennis Marcus takes us behind the scenes where Cruden’s 3D content artists create detailed track and road graphics.
We’re often asked how Cruden goes about building the detailed 3D track environments that run in our driver-in-the-loop (DIL) simulators. The short answer is: it’s not as easy as you think!
Naturally, if we are writing about (race) tracks, then this article is going to be more relevant to the motorsport world, where we supply track models to race teams, championships and motorsport engineering service providers.
In the automotive industry, our traditional clients use their simulators for chassis development, for which the same type of content as in motorsport is typically used, ranging from the Nürburgring Nordschleife to various proving grounds. The rapidly growing use of driving simulators for autonomous vehicle development and in HMI studies demands a different type of content; where a typical race track model has a length of about 5 km and requires millimetre-precise shape and position of kerbstones, our automotive customers require hundreds of kilometres of public roads with real world crossings, traffic signs, road marks, urban areas plus country roads and highways.
Simulators at the highest level of motorsport are used as an engineering tool, they’re not primarily there to train the drivers. Teams use simulators to help engineers figure out the best setup direction for the car on a particular track. To achieve this goal, the computer models of the race car, and the tire models within, must be fed high-quality road definition from LiDAR scans.
The driver needs to see the road, but the car needs to feel it. The road input to the tire models based on which the forces are calculated must be based on actual measurements rather than an artificial road surface.
When Cruden creates new racetrack content for the simulator, or if an automotive customer asks us to make an accurate 3D model of a public road or proving ground track, we typically start with a LiDAR scan. A day will be spent in a car equipped with LiDAR, IMU and GPS sensors to record every detail not only of the track surface, but of the surroundings as well. For a race-car driver, it’s important for example to include a tree or a marshalling post that they use as a braking marker, in the exact same position as it is on the real track. The things closest to the driver must be reproduced in the highest possible level of detail to make the DIL experience as immersive as possible.
When the scan is complete, the LiDAR, IMU and GPS data is processed to create a set of points that are representative of the actual road surface and surroundings to an accuracy of a couple of millimeters. The data exists in two forms: the exceptionally detailed physics layer, which provides the interface plane between the virtual car and world that is not rendered but processed separately on a different computer at runtime, and the visual database, which is the data processed to create the objects the driver sees. To create the latter, we load the point cloud file into 3D design software to guide our highly skilled 3D artists in placing each object at exactly the right location, supported by video footage and photography that we also shoot on the day of the scan.
It can take four to six weeks to complete the 3D design process, although sometimes it is inevitable to do it in a much shorter timeframe, such as for a Formula E street circuit. Cruden does LiDAR scanning for the FIA Formula E Championship and creates 3D graphics for several Formula E teams. For this series, we do the LiDAR track scan as late as possible, to ensure it’s as accurate and relevant as we can make it.
For automotive driving simulators, creating content based on LiDAR scan data is not always a requirement. For some experiments, there is a benefit to creating a virtual road network that enables a series of specific scenarios to be tested. In the automotive world, accurate 3D content based on point cloud data is used to reproduce issues that were detected with a real car on a real road in the simulator. When the engineers are able to reproduce an unwanted behavior of a system, they can use the simulator to test and validate their technical solution.
When it comes to building track or road graphics, it is tempting to employ a recent design school graduate, trained in building game graphics. But building 3D graphics to run in an engineering simulator is a very different discipline. Many 3D graphics designers have no understanding of vehicle models or tire models. Building graphics and dealing with the LiDAR layer underneath is difficult for these guys and in our experience, they often struggle. As a result, the customer gets something that looks nice but doesn’t work in the simulator because the designer has no understanding of engineering simulators.
At Cruden, we understand simulators and through our customers, we get a lot of feedback from professional race-car drivers and simulator drivers. Based on that, we know that we must focus on the part of the visuals that has an impact on the experience of the driver.
If you build a 3D model of the Formula E street track in Paris, the Eiffel Tower is not important. What is important is to have all the right details on the road surface and to place the trackside advertising banners in the right position. We can achieve an incredible level of detail in the areas where it counts, thanks in part to recent advances in graphics card performance and render technology – a topic we’ll explore in a future article.
Look out for our next article on the differences between visuals for engineering-centric simulations, such as sensor simulations for ADAS development, and human-centric simulations for fully immersing driver test subjects.
For more information, please contact Dennis Marcus via firstname.lastname@example.org or on +31 20 707 4646.
If you think a colleague would be interested in receiving the articles in this series, you can sign them up to our newsletters here.
Other articles in the series:
Links to subsequent articles will be added below as they are published.