[Game Changer] Simulator startup aims to make self-driving cars safe
Published: 16 Jul. 2023, 06:00
Updated: 19 Jul. 2023, 10:22
To develop autonomous vehicles (AVs), one needs to think through multiple possible scenarios on the road where all sorts of uncertainties and dangers lie.
And that is no easy feat, or a cheap one.
This is where autonomous driving simulation company Morai comes in with its Morai SIM-Drive simulator.
“People get behind the wheel to go where they want whenever they want, come rain or snow, and regardless of whether the temperature is high or low,” Jung Ji-won, CEO and co-founder of Morai, said.
“That makes testing [AVs] difficult because there are so many scenarios that need to be considered,” he added.
Founded in 2018, Morai has worked with leading AV developers both at home and abroad, including AWS, Nvidia, SoftBank and Ansys, as well as Hyundai Motor, Samsung Electronics, Naver Cloud and Hanwha Aerospace.
Carmakers worldwide, including Tesla, repeatedly hit the headlines with accidents involving self-driving advanced driver assistance systems (ADAS) features, partly because of the complexity of the technology derived from the tremendous amount of possible variables that need to be considered.
There is no such thing as too much safety when it comes to autonomous driving — and that equals an opportunity.
The global driving simulation market size is expected to be near $2.8 billion by 2030, up 82.8 percent from 2021, according to market tracker Verified Market Research. Meanwhile, the ADAS market is forecast to reach $65.1 billion by 2030, from last year's $30.9 billion.
Morai provides its driving simulation software to AV developers and component suppliers, creating a safe and cost-effective environment to repeatedly test self-driving systems in a virtual space with real-life-like settings.
One of the Morai SIM simulator’s forte over its competitors is its mapping technology, which automatically creates a digital twin map of real-world environments based on existing geographical information, the company said.
"Other companies do not have such a technology, because those programs require manual involvement in updating the map," Jung said, explaining that whenever there are changes made to the actual geographical environment, Morai's mapping system linked with real-life data will automatically incorporate the changes into its 3-D map.
Morai's simulation platform offers a virtual testing space for not only AVs (Morai SIM Drive) but also aircraft such as urban air mobility (UAM) vehicles and drones (Morai SIM Air), and plans to expand the application into robotics (Morai SIM Robotics).
The company completed a Series B funding round earlier last year and has so far raised 30 billion won ($23.7 million) from major investors including Naver, Hyundai Motor, and Kakao Ventures.
Morai set up a U.S. entity in San Francisco in 2020, which mostly focuses on projects involving UAM, and plans to secure a foothold in Germany.
The Korea JoongAng Daily sat down with Jung for an interview at Morai's headquarters in Gangnam District, southern Seoul, to hear about the company's journey so far and plans for further expansion across borders.
This interview has been edited for length and clarity.
Q. What motivated you to enter the autonomous driving simulation market?
A. When our chief technology officer and co-founder Hong Jun and I were in a lab at KAIST together, we mainly focused on testing and verifying the performance of self-driving cars, drones and robots.
There have been a number of accidents on the road for other companies developing autonomous driving technologies, such as Tesla, and we saw a lot of accidents or near accidents during our time in the lab as well.
Running such tests on the actual road is not only dangerous and inefficient, but also very costly. So we thought that developing a driving simulator would improve the process efficiency with better safety and convenience.
Testing a self-driving car also involves various pieces of equipment, such as sensors, and makes the process more difficult, leading to many unexpected and sometimes dangerous situations.
A car is supposed to enable the driver to conveniently get to certain places or move certain things regardless of the season or weather, or time. It is therefore crucial to run the tests in various environments.
How is artificial intelligence (AI) incorporated into your Morai SIM technology?
The company’s name, Morai, can be interpreted in two different ways: It is a combination of ‘Mobility, Research, and AI,’ but also reads as ‘More AI,’ indicating that we provide a large number of AI datasets.
Creating data via on-road tests in various environments can be challenging, and even if we secure the data, it needs to be labeled by hundreds of people for machine learning use.
Our simulation tool, however, collects and creates data in a labeled format, which can be used as training data for machine learning or a verifying tool for a trained self-driving system’s performance.
How is Morai SIM Cloud, which was first introduced at CES 2022 last year, different from your previous services?
Previous simulators usually categorized vehicle components in largely three segments: Chassis, body and powertrain. Such hardware equipment, though not simple, did not require that much simulation testing.
However, when it comes to autonomous driving, we now have to consider a huge number of environmental variables.
Running a number of tests in a short period of time requires a tremendous amount of computing resources in a cloud environment. Therefore, Morai SIM Cloud utilizes a maximum amount of resources on the cloud to conduct as many tests on various scenarios as possible.
That means that we can use a larger amount of computing resources in a more efficient manner. For clients or developers, they wouldn’t need to spend money on establishing infrastructure and instead use the existing resources from other cloud service providers such as Naver Cloud.
Automatic mapping technology is considered one of Morai’s biggest strengths over its competitors. Could you explain this further?
For existing simulators before Morai, it was nearly impossible to completely recreate a virtual testing space based on the actual real-life environment, because someone had to manually draw a map of a space measuring hundreds or thousands of square kilometers, which also couldn't be free of human errors.
We developed our mapping technology using high-precision map data provided by the Ministry of Land, Infrastructure and Transport. The map data has been made available to the public since 2019, and we began research and develop a way to virtualize the vast amount of geographic information, using AI technologies and logic.
To create a virtual version of cities outside Korea, such as Las Vegas, we mainly worked with private mapping companies and sometimes with the local government, which was the case for Singapore.
Morai SIM Drive was granted the ISO 26262 safety standard certification for electrical and electronic systems in March. What does this mean?
Because the ISO 26262 certification guarantees a product’s credibility and safety, a component manufacturer needs to submit a qualification of the tool it used during the product development in order to be granted the certification.
Having the ISO certification thus helps us immensely in clinching supply deals for our clients, especially component manufacturers.
Without a tool qualification, we need to provide too many documents, so being granted the ISO certification proves our trustworthiness to potential clients and simplifies the process.
We are currently working with TÜV SÜD, a German certification authority, to secure additional qualifications, and plan to further secure even more validations and qualifications in the field.
Could you tell us more about Morai’s collaboration with local and global companies?
We work most closely with Ansys among overseas companies. Because Ansys has many tools that can be used to interpret hardware such as sensors, signals and systems, we pair the company’s technology with our simulator, which creates digital twins to strengthen our competitiveness.
With AWS, we run our Morai SIM Cloud on the AWS’s cloud service and are working with the company to further secure users in the AWS cloud ecosystem. Moreover, we are also in technological cooperation with AWS with the goal of utilizing its technologies to evaluate and validate autonomous driving systems.
With Nvidia, we are using its hardware products, such as graphic processing units, to run the Morai SIM program. And because Nvidia has Omniverse, a virtual space creation platform, we are also collaborating with the company on projects related to Omniverse.
Naver, which is both our client and investor, is using our simulator to develop autonomous driving technology, and we are using the mapping data from Naver’s ARC — AI-Robot-Cloud — system.
We are also working with Naver Labs, or Naver Cloud, for government-backed research and development projects, and are collaborating on a joint project in Japan as well.
What is your plan for expanding the business and the application of your service?
In terms of technological development, the most important thing for us is automatizing a bigger number of technologies, and therefore enabling automatic creation of more diverse environments and scenarios for the users.
And one area that combines autonomous driving, UAM and robotics is defense.
We are currently working on many defense projects. For example, there is this thing called manned-unmanned teaming, in which manually-controlled defense systems and autonomous weapons conduct missions together. Morai is developing a war simulation platform to integrate our technologies involving cars, UAMs, and robotics.
K-defense is on the rise as of late, so we are finding new applications in the field by integrating our technologies into one system.
BY SHIN HA-NEE [shin.hanee@joongang.co.kr]
with the Korea JoongAng Daily
To write comments, please log in to one of the accounts.
Standards Board Policy (0/250자)