VTD helps Waylancer with fast, concurrent simulations

Shanghai Waylancer Automobile Technology Co., Ltd was incubated by the SAIC Group Innovation Fund in 2017. Waylancer provides development, software, and big data product services for intelligent driving simulation testing. The company has provided a cloud simulation testing platform for companies like Lotus Cars, Changan Auto, and GAC Grou.

Challenge: Help clients embrace ADAS/AD 

Waylancer wanted to help its OEM clients embrace the industry’s move towards advanced driver assistance systems/automated driving systems (ADAS/AD). To do this, the company knew it needed to help its clients overcome many challenges.

Increasing complexity without increased testing time

Over the years, ADAS and autonomous systems have exponentially become more complex. In the past, features like adaptive cruise control or blind spot warning may use a few sensors. Still, as the industry has evolved, the number of sensors in vehicles is increasing — in some cases, up to 20 sensors, with information that needs to be “fused” with input from other sensors to make a single “act” decision. OEMs are gathering multiple data points from various sensors but have the same amount of time to perform testing scenarios.

“One good example is a camera, which is adept at detecting different types of objects but performs poorly when trying to detect the object’s distance from the vehicle. Simultaneously, a radar sensor is terrible at detecting an object’s type but excels at determining an object’s distance from the vehicle,” explained Dr. Furea Shirai, Hexagon’s product marketing manager, autonomous and cloud.

“By combining data from both sensors, the vehicle can generate a better perception. But the manufacturer must still test both sensors together — and as they combine information gathered from sensors and determine test scenarios, it’s easy to see how the number of scenarios customers must test and validate grows exponentially. Manufacturers have the same amount of test and development time in which they have to test ten times the amount of scenarios,” she said.

Reliance on physical testing

In many manufacturing settings, OEMs perform most of their testing scenarios in the physical world, putting prototypes through a wide variety of potential driving scenarios on proving grounds to see how the vehicle performs. However, this is an expensive, time-consuming process that requires the prototype to be available for the OEM to perform any testing. Any potential issues found in testing happen late in the design and development process, meaning any changes that need to be made will be costlier than if they had been previously discovered.

Difficulty testing everything

While the number of test cases has grown vastly, it’s practically impossible for manufacturers to test everything in road testing because it’s expensive and time-consuming. Manufacturers do have to meet specific standards — like the National Highway Traffic Safety Administration’s New Car Assessment Program (NCAP) and ISO26262, which defines the international functional safety standard for road vehicle electrical and electronic systems development — but they must test beyond those standards to cover the wide variety of use cases associated with real-world driving.

“Because there is no standard for evaluation, we can’t develop an excellent evaluation algorithm that can evaluate all ADAS/ AD algorithms from customers,” said Shirai.

Solution: Virtual Test Drive (VTD) facilitates fast, concurrent simulation testing

Waylancer began evaluating test-driving toolkits, ultimately choosing Hexagon’s Virtual Test Drive (VTD) in 2018 due to its cloud-based architecture, flexible scenariobuilding capabilities, and excellent scenario-rendering capabilities. Additionally, VTD is one of the leading simulation environments in China and is well-known and well-established among Chinese automotive manufacturers, making it a natural choice.

Waylancer provides its customers with a turn-key cloud simulation testing platform, allowing customers to perform high concurrency simulation in the cloud. The docker version of VTD, combined with Waylancer’s cloud simulation testing platform, helped the company realise high-concurrency simulation in the cloud, greatly accelerating the testing process. Its cloud simulation testing platform supports high-concurrency simulation, allowing the company to test multiple scenarios simultaneously, accelerating algorithm iteration and saving costs.

Client experiences

One of Waylancer’s clients, Lotus Cars, uses Waylancer’s cloud simulation testing platform with 100 VTD dockers. This allows Lotus to simultaneously simulate 100 scenarios and output evaluation reports, providing insightful action items Lotus could use to improve its AD algorithm.

Over the course of 2,127 hours of simulations, VTD simulated 12 million scenarios, with each scenario taking approximately one minute. Lotus pulled some scenarios from the NCAP, but others came from Lotus’ existing scenario library.

Additionally, Waylancer provided a cloud simulation testing platform for another Chinese auto manufacturer with 50 VTD dockers. The platform combined the open-source Simulation of Urban Mobility (SUMO) traffic simulator package to perform intelligent traffic flow simulation, providing more realistic traffic flow data for AD algorithms. It also supported a combination with Ansys AVxcelerate sensors for perception simulation, output the raw data of cameras, lidars, and millimetre-wave radars, and provided a training set for perception algorithms.

Results: 80 times faster than HIL

Future plans

In the future, Waylancer plans to develop two platforms with VTD:

It first intends to develop a cloud-based perception simulation platform that supports the online configuration of sensors (such as cameras, lidars, and millimetre-wave radars). In addition, it will support lidar and millimetre-wave radar simulation using VTD’s Optix plug-in and output point cloud data. It will support camera simulation through VTD’s IG module, outputting fisheye, semantic segmentation, and other camera images.

A cloud-based traffic flow simulation platform that will combine with PTV VISSIM multimodal traffic simulation software for co-simulation of traffic flow, allowing users to configure traffic flow parameters online, such as traffic flow and traffic density. The traffic flow simulation platform will include a variety of built-in traffic flow models, such as following models, lane change models, etc.