Image courtesy of AV Simulation

Multi-purpose car simulation environment gets a boost from Unreal Engine

Sébastien Lozé |
24 mai 2021
As the automotive industry rapidly transitions from traditional cars to advanced driver assist systems (ADAS) and autonomous vehicles, the driving simulation industry must transform at the same pace. The testing of such vehicles is a huge task, where detailed environments, enormous data sets, sensors, and a number of other factors need to come together to produce a useful result. 

Driver-in-the-loop (DIL) driving simulators, with the driver seated inside a physical simulator, have been a staple of the automotive industry for decades—manufacturers rely on them to improve how the driver, the car, and the road interact. Today, self-driving and assisted-driving cars require their own virtual simulators to replace thousands of miles on real roads with billions of digital miles in a wide variety of road conditions.

Paris-based AVSimulation, a joint venture between simulator company Oktal Sydac, automaker Groupe Renault, and CAD software maker Dassault Systèmes, was formed to tackle these problems by offering a full suite of automotive testing tools, including driving simulators and support systems. In addition to their multi-purpose simulation software SCANeR, the company currently builds some of the largest physical driving simulators in the world, with both Renault and BMW as clients. 
 

“The tools we offer are evolving to tackle the huge amount of data required to test and validate vehicles that are becoming more and more advanced,” says Emmanuel Chevrier, CEO at AVSimulation. “Our vision is to provide all the tools required by customers and other services, to help them accelerate in the race towards safer and more autonomous vehicles.”

To answer all these needs, AVSimulation recently switched to Unreal Engine as SCANeR’s 3D engine. The result is a powerful solution that keeps SCANeR’s original promise for modularity, and accepts data from a wide variety of other applications. The software also includes realistic visuals, an easy asset import and access workflow, straightforward GIS import, and a host of other features.
Image courtesy of AV Simulation

SCANeR studio: second revolution 

SCANeR studio, a suite of products at the core of AVSimulation’s offerings, is a generalized platform for all phases of automotive testing including the use of software-in-the-loop (SIL) and hardware-in-the-loop (HIL) simulators.

SCANeR was invented by Professor Andras Kemeny, Expert Leader in Simulation and Virtual Reality at Renault. AVSimulation’s CTO Guillaume Millet revolutionized the product more than 10 years ago, and rebuilt it again around Unreal Engine with a team of engineers. The rebuild was required to tackle the incredible complexity, realism, and size of the simulations auto manufacturers now require, and also to interface with the many different embedded systems, software platforms, and bespoke systems individual clients may bring to the table.

“You can bring your existing sim modules, such as headlights or vehicle dynamics, or use the out-of-the-box SCANeR solutions,” says Chevrier. “We want to interface with all other products, be they IT or industrial.”
Image courtesy of AV Simulation
He explains that while various “in-the-loop” technologies are required to fully test an autonomous vehicle, AVSimulation’s customers would like to use the same software for DIL, SIL, and HIL phases. He adds that the software can also support a vehicle-in-the-loop (VIL) phase, which is usually accomplished through virtual reality simulation. 

“The beauty of SCANeR is that it can be used throughout these phases,” Chevrier says. 

SCANeR also offers different levels of complexity at the UI level to accommodate customers who want a plug-and-play solution as well as those who want to dive into programming and the more technical aspects of Unreal Engine.

“The standard user of SCANeR just wants better rendering and better images,” says Thomas Nguyen That, Automotive Domain Director at AVSimulation. “At the same time, we have advanced users that want to be able to customize the rendering, so we have to also provide that entry for them to Unreal Editor.”
Image courtesy of AV Simulation
Porting the system to Unreal Engine helped AVSimulation make the system more modular and scalable, offering different solutions for specific roles or approaches to simulation. A user can remove any module such as vehicle dynamics or headlights, and replace it with another module from an outside vendor. To familiarize their users with the SCANeR next generation, AVSimulations has created a video to show how to set up SCANeR and Unreal Engine together.

Realism with Unreal Engine

As part of its commitment to realism, SCANeR includes tools for building a virtual world, providing everything a simulation environment might need: roads, landscapes, vehicle dynamics, traffic, sensors, headlights, weather conditions, data from real or virtual drivers, and scenario scripting.
Image courtesy of AV Simulation
To get to the required level of realism, SCANeR must generate camera data an order of magnitude more accurate and convincing than what the human eye needs. A few choice vertical lines might say ‘rain’ to your eye, but a computer, lacking the interpretive abilities of human beings, requires photorealistic rain in order to understand it. 

“One key reason to change the technology was to target photorealism,” says Millet. “With Unreal Engine, we get access to new technology like ray tracing, we can do physics-based rendering simulations, and we are able to provide better input data to sensors and also to humans.”
Image courtesy of AV Simulation
Another advantage of porting SCANeR to the Unreal Engine platform is that it gives their customers access to the Datasmith suite of import tools. “Datasmith is very useful for us, especially at runtime, to achieve almost the same workflow the users are using with simulation,” says Millet. 

And last but not least, the team also likes the fact that the Unreal Marketplace provides reusable assets that their customers can use to build their 3D environments quickly. “The ecosystem was another driver,” says Chevrier. “Value for money was definitely an aspect that was appealing to us in the choice of Unreal.”
Image courtesy of AV Simulation

Working with other technologies

With regard to scalability, AVSimulation is able to use SCANeR as the backbone for huge simulators on rails that give users the opportunity to experience road infrastructure before the roads themselves are built. The clients for these systems include Renault and KEX of South Korea.

The team also built two large simulators-on-rails systems for BMW: one for an urban environment, and another for simulation of evasive manoeuvres, full braking, and hard acceleration. The latter, explains Chevrier, is necessary because “on some German highways, there is no speed limit.”
Image courtesy of AV Simulation
Such simulators require an extraordinary level of power. “We had to find all these very niche players from carbon to speed train technology, and aggregate all these crazy technologies to build this simulator,” he continues. “There are also a lot of electronics and embedded software required. Because at the end of the day, these are really just huge robots.”

Because of the flexibility Unreal Engine brings to SCANeR, the team was able to use the software for these projects as well, fusing the many technologies required to deliver a single, complete system.

Chevrier admits that AVSimulation’s decision to convert SCANeR to the Unreal Engine platform was as much about Epic Games as a company as it was about the engine itself. “One of the things we look for in a partner is commitment,” he says. “Epic Games is a gaming company, but what we found is a strong long-term commitment to this space.

“This is a race. The customers must race to find the right asset, and the right software if they want to be the first to put their autonomous vehicles on the market. Unreal Engine brings a competitive advantage for any simulation tool, be it a driving simulator or a set of computers running a massive simulation.”
Image courtesy of AV Simulation

    Let’s talk!

    Interested in finding out how you could unleash Unreal Engine’s potential for simulation?
    Get in touch to start that conversation.