Just like any other vehicle destined for public use, autonomous vehicles (AVs) require thorough testing—and lots of it. Not all of that can be carried out in the open, however. Strict regulations dictate how prototypes can operate amid normal traffic, with various hoops to jump through before receiving approval. Even then, so-called corner cases—rare incidents that could prove problematic to an AV—are not ideal when considering the potential damage that could be caused.
At off-road test facilities away from the public eye, that is not the case. Test dummies can take the place of pedestrians, crashes will not incur any lawsuits, and corner-cases can be repeated over and over. As such, anything from abandoned military facilities to derelict manufacturing sites have found a new lease of life as AV proving grounds, letting developers off the leash to put prototypes through their paces—safe in the knowledge that any risk is primarily limited to financial loss at worst.
There are hundreds of vehicle testing grounds around the world, but very few that can replicate the complex driving environments an AV will face on a daily basis. It is not simply about honing attributes such as noise, vibration and harshness (NVH). These cars need to learn how to interact with complicated intersections and pedestrian crossings, and how to react to other road users—particularly vulnerable actors such as cyclists. As would be expected, the necessary scale and subsequent cost of such a facility means they remain relatively few and far between.
Proving grounds
The GoMentum Station in Concord, California, is a staggering 2,100-acre abandoned naval base that has been used as a test bed by the likes of Honda, Toyota and Lyft. Paul Cummings, Group Lead for Systems Integration, Automated Vehicle Research at Honda Research Institute USA, has described the base as “an ideal proving ground” due to its controlled environment, “which can be continuously modified to represent a wide array of settings that an automated vehicle must navigate, especially for urban operation.”
There are hundreds of vehicle testing grounds around the world, but very few that can replicate the complex driving environments an AV will face on a daily basis
Both Waymo and Uber have their own test facilities in California and Pittsburgh respectively, which act as replica towns. Then there is CETRAN, the Centre of Excellence for Testing & Research of AVs, a 4.5-acre facility operated by the Nanyang Technological University in Singapore and supported by the city-state’s Land Transport Authority. It was built from the ground up in 2017, with mock roads, intersections, and even ‘skyscrapers’ to replicate potential cell-signal interference. A small section is also outfitted with a rain and flood simulator.
Opened in 2017, K-City is a 79-acre mock town situated roughly an hour from Seoul. It is South Korea’s primary private test bed for AVs. In the UK, the RACE facility located in Oxfordshire houses a 10km network of tracks to test both connected and autonomous vehicle technologies. A similar facility has been established by the Research Institutes of Sweden and Chalmers University in Gothenburg; the AstaZero facility contains a variety of test environments, including rural roads, multi-lane roads, high speed areas and city driving. Transpolis, a dedicated ‘urban mobility lab’ in Lyon, France, is also a test bed for AVs, among other technologies. Available to developers is a one-kilometer highway, a ring road, a country road, and a fake city of 30 hectares where traffic lights, pedestrian crossings and bus stops help to replicate urban driving.
Motor City becomes mobility central
Michigan in particular has become something of a hot bed for such facilities. Consider the University of Michigan’s Ann Arbor-based Mcity; the 4000-acre Chelsea Proving Grounds, which has received a US$30m investment from FCA; the Toyota Research Institute’s 60-acre site at the Michigan Technical Resource Park; and the American Center for Mobility (ACM).
The latter is a 500-acre site adjacent to Willow Run, a former manufacturing park in Southeast Michigan best known for making aircraft during the Second World War. Having served under the ownership of both Ford and General Motors in the past, its sprawling grounds today house the ACM, a joint initiative developed by a handful of players including the state Department of Transportation (DoT), the University of Michigan and the Michigan Economic Development Corporation.
We act as a bit of a matchmaker—it is a neutral meeting point to do collaborative research without a formal agreement
Like most AV proving grounds, the idea is to foster the development and deployment of future transportation systems. It works with automakers, Tier 1 suppliers and a wide array of infrastructure companies. Relevant players also have the option to invest and sit on its Industry Advisory Board. “But they don’t just plunk down dollars and then come in and test,” advised Jeff Rupp, Chief Technical Officer at ACM. “There are programmes, experiments and research projects in which they want to be involved, and we can help them deliver those types of projects. We act as a bit of a matchmaker—it is a neutral meeting point to do collaborative research without a formal agreement.”
The benefit of such a large playground is that testing can essentially be carried out in realistic scenarios, rather than having compact test cases one after the other. Compared to other facilities in the area, it is one of the larger stretches of land on offer. “Mcity has a lot of great features, they’re just very densely packed,” explained Rupp. “If you’re driving from one end of the site to the other, you might experience ten different unique things. But in the real world different types of roads are not that densely packed. If you just want to experiment on one element, you only get a short run at it.”
Sister act
As it happens, Mcity’s Director, Huei Peng, sits on the ACM Board of Directors. “Their staff and our staff meet regularly,” noted Rupp. The two entities are currently working together to find projects that can “graduate” from Mcity to ACM. “We’re working on a research project proposal right now involving data acquisition systems and data management,” said Rupp, “because there are massive amounts of data being generated by AVs, and you need to be able to ingest all of that and make it useful.”
If you want to replicate real world scale, you need to come to a larger facility. We have areas where you can hit higher speeds for longer periods of time—it’s more naturalistic in terms of driving
As one would assume, creating realistic driving environments in which to hone AVs requires space. Of the ACM’s 500 acres, vehicle test environments take up around 300 acres or so. By comparison, Mcity has about 16 acres of roads and traffic infrastructure. “If you want to replicate real world scale, you need to come to a larger facility,” continued Rupp. “We have areas where you can hit higher speeds for longer periods of time—it’s more naturalistic in terms of driving.”
More compact facilities such as Mcity, he says, are more focussed on performing initial research and development activities, testing concepts and trialling innovative approaches. The next step is usually to expand that programme. “When you’re ready to go larger scale or prove out commercial readiness, you need a different type of facility,” said Rupp. That said, he is quick to underline that there is no competitive dynamic between ACM and Mcity. “We are really more of a sister collaborative; we’re certainly not competitive in any way because our purpose and intent is very different.”
All for one
Previously a veteran of 15 years at Ford Motor Company, Rupp reflected on how his position in the industry has changed. Rather than helping one automaker move forward, he is now part of a wider entity assisting the entire industry in its shared pursuit of future mobility solutions—and not only automakers, but also Tier 1s and infrastructure providers.
“It is probably a better use of my time, and a better way to have an impact as an individual,” he mused. “So many people are hurt or killed in crashes, it is a huge problem. When a human driver is on his or her game, I’m not sure that a machine can perform better unless it has access to information a human does not. But humans are fallible: they make mistakes and get distracted, and it is not possible to maintain the level of attention that is required. That fallibility indicates there is a huge opportunity to make a change.”
Humans are fallible: they make mistakes and get distracted, and it is not possible to maintain the level of attention that is required. That fallibility indicates there is a huge opportunity to make a change
Safety is a driving factor behind the billions of dollars being pumped into driverless vehicle development, but there are other societal issues that must also be considered. Not everyone has the same level of access to transportation, and vehicles must be designed and deployed with the needs of all riders in mind.
“We want to do whatever we can to help the industry move forward as quickly as possible to achieve all of these great benefits in terms of safety, fuel economy, accessibility and time efficiency,” concluded Rupp. “We are not picking winners and losers at ACM, we are trying to help everybody progress. We will learn, they will learn, and we can accelerate everybody—we’re all in it together.”
This article appeared in the Q3 2019 issue of M:bility | Magazine. Follow this link to download the full issue.
