Self-driving Test Sites Use Virtual Pedestrians to Streamline Development

author: Michael Cheng   

For car manufacturers developing driverless cars, recreating critical road scenarios in testing sites can be a tedious experience. Moving around traditional test dummies (like the ones deployed in collision labs by safety researchers) and manually setting up road obstacles is extremely time consuming, as well as costly to implement.

A solution to such challenges involves the utilization of virtual pedestrians and obstacles. Swedish automaker Volvo and established truck manufacturer Scania are currently leading this emerging trend, which is being leveraged at AstaZero, a private test track that host trials for developers of self-driving vehicles. The cutting-edge technology is also applicable to testing autonomous trucks, delivery bots and drones.

Deploying Virtual Pedestrians and Obstacles

Virtual pedestrians are not designed to replace trials on public roads. However, such techniques can streamline the process by reducing a range of manual aspects at the early stages of testing autonomous vehicles. For instance, on a private test site, developers may deploy virtual pedestrians crossing at a busy intersection, while a self-driving vehicle attempts to maneuver through the scenario safely.

If real model figures were used, individuals would have to reset the test dummies after every crossing. Manually resetting obstacles is also less accurate than using programs that are capable recreate the same virtual scenario in a loop, without compromising placement, speed of movement and pedestrian-like behaviors.

"In this kind of environment we're able to test more or less what we need for public roads later on," highlighted Lars-Gunnar Hedström, Engineering Director at Scania's Connected and Autonomous Systems, during an interview with Reuters.

"We have the possibility to be out on customer sites and run real operations much earlier, which is a big difference."

When it comes to testing autonomous cars in complex scenarios, such as merging lanes, blind corners and exposure to rough weather, virtual environments could make driverless units less prone to physical damage from potential collisions. From another perspective, virtual obstacles would ensure human developers are safe riding inside the vehicles during trials.

Moreover, businesses conducting automotive testing with autonomous cars may also consider that repair stemming from collisions with real obstacles is greatly decreased, when using virtual alternatives on private tracks.

Safety First

Ultimately, the use of virtual pedestrians to enhance testing for driverless cars widely proliferated from responses to an accident involving an Uber self-driving vehicle and a pedestrian crossing the street on public roads. Both experts and consumers voiced their concerns about making trials for autonomous vehicles safer.

In a move to prevent such accidents from occurring, a handful of car manufacturers are turning to virtual testing solutions to hone the capabilities of their platforms before conducting trials on public roads. Regulators have also stepped in to help update various testing protocols. All of these actions point to a maturing autonomous-vehicle industry that is set to deliver promises of a functional, safe and exciting driverless future.

"Autonomous technology has the potential to... reduce the number of accidents. That's something we need to work with jointly in this industry," said Robert Falck, CEO of Einride.

Michael Cheng

Michael Cheng is a legal editor and technical writer with publications for Blackberry ISHN Magazine Houzz and Payment Week. He specializes in technology business and digesting hard data. Outside of work Michael likes to train for marathons spend time with his daughter and explore new places.