Mcity unveils first open-source digital twin for mobility systems testing

The University of Michigan’s autonomous vehicle testing facility, Mcity, recently released its digital twin, a virtual replication of the physical facility at Mcity. The replication will allow researchers to test their algorithms without booking time at the physical site.

Mcity began operations in 2015 with the world’s first testing facility dedicated to autonomous vehicles. The facility spans 32 acres on the University’s North Campus. 

The virtual representation of the facility uses real-time data to simulate the physical environment. The Mcity digital twin is built on top of CARLA, the largest open-source simulator for the testing of autonomous vehicles and sensors.

In an interview with The Michigan Daily, Mcity software engineer Darian Hogue said the software took nearly a year to build and was developed with support from the National Science Foundation. He explained creating an accurate digital twin required a long process of gathering real-world data from the physical test site.

“You take the position of the physical vehicle and you basically load that into CARLA or a digital twin, and you kind of do a one-on-one comparison to see if the physical vehicle is being displayed correctly in the digital environment,” Hogue said. 

Rackham student Zhijie Qiao, who is pursuing a doctoral degree in the College of Engineering and is involved with research at the Michigan Traffic Lab, told The Daily the digital twin allows researchers to test and debug software more quickly. 

“On the track, it’s hard for you to actually develop code and just debug what’s wrong,” Qiao said. “But instead, if you have a digital twin … a lot of that work can be done in advance. So when you actually go to the track, you can get straight to the point.” 

Mcity Director Henry Liu explained in an interview with The Daily that while the digital twin has many uses, it is still good to utilize both the digital twin and the facility together in some cases. 

“You can’t really replicate everything and so there’s a discrepancy in between them,” Liu said. There’s what we call the simulation-to-reality-gap. After you test in the digital twin, you want to verify your test in the physical world, so that you have more confidence.”

For example, when it is unsafe to test certain activities such as  pedestrian traffic in real life, researchers can use a proxy object — a physical object meant to represent a pedestrian. Liu said that using the digital twin, the proxy object can be made to represent a person. 

“It’s not only just the physical world, but the digital object is also infused into (the automated vehicles’) images, the video, and so they have to react to it,” Liu said. “So this is one example of the use of the digital world in combination with the physical world to create what we call a mixed-reality testing environment.”

While grants are available for researchers hoping to make use of the physical track, utilizing a test site like Mcity is often costly. Hogue hopes the digital twin will help expand access to the facility to those who might not otherwise be able to use it. 

“We know that there are other digital twins out there for use, but usually it costs a pretty penny to get access to that,” Hogue said. “Since we’re based at the University, we were thinking ‘okay, how can the students access this?’ And the best way to do that is to make it open source.”

Daily Staff Reporter Amanda Venclovaite Pirani can be reached at amandavp@umich.edu.

The post Mcity unveils first open-source digital twin for mobility systems testing appeared first on The Michigan Daily.