Argo AI is one of several tech start-ups testing autonomous vehicles on public roads
China’s focus on public infrastructure puts it leagues ahead in the global race to level four autonomy
When is an autonomous car not an autonomous car? When it relies on feedback and information from infrastructure and other vehicles around it.
As the UK works to unlock what the government says could be a £42 billion industry moving goods and people around in driverless vehicles by 2035, the worry is that we’re asking too much of the vehicle and its on-board technology and not investing enough in the support network that could better facilitate its rollout.
“The dictionary definition for autonomous is not relying on anything else,” Peter Stoker, chief engineer for connected and autonomous vehicles at global automotive testing company UTAC, said. “Some in this business believe we shouldn’t rely on, for example, live data because sooner or later it’s not going to be there.”
However, as the goal of full autonomy looks ever more distant, we might need connected infrastructure to give it a helping shove. Stoker draws an analogy with laptops. “They were a bit clunky to start with but they really took off when you connected them all together,” he said.
Infrastructure helpful to autonomous vehicles includes tech like roadside data transmitters used to share information with the car. It also includes cameras or lidar sensors fixed on poles to better locate vehicles. The idea is that you create corridors such as motorways or urban zones to enable what’s known as level four autonomous driving, which is completely hands-free but restricted to certain areas.
That’s happening right now in China, a country that plans to invest $300 billion (£259bn) between now and 2040 in what’s described as ‘vehicle infrastructure cooperation (VIC),’ the research arm of UBS bank forecast in a recent report. “VIC should drive faster autonomous implementation as it removes vehicle-side technological bottlenecks,” the report said.
The analysis shows the prizes we could win if we followed a similar path and also reveals the stumbling blocks that, while solvable for China, could be insurmountable for the UK, given our less centralised planning.
China’s rollout already includes upgrading motorways to a level of smartness that far exceeds our efforts. For example, the recently upgraded Shanghai-Hangzhou-Ningbo Expressway includes embedded sensors that connect to a cloud computing network to monitor vehicles in real time. Sections even come with embedded wireless charging for EVs.
The mission is partly to improve traffic speeds, with the data infrastructure eventually allowing the mass control of driverless cars to speeds of 75mph, much higher than current averages, by commanding vehicles to brake and accelerate at the same time.
Initially, the first users of autonomous-enabling roadside infrastructure in China will be lorries, UBS predicts, which are seen as delivering the biggest cost benefits for self-driving because they eliminate the drivers, which currently account for around a quarter of haulage costs, UBS estimates. Essentially, the corridors would be road versions of railways, with the autonomous lorries offloading to logistics centres and manned vehicles carrying out the final-mile delivery.
Right now, the burden for delivering autonomous vehicles lies with specialist companies such as Cruise, Argo AI, Waymo or, on the trucking side, Aurora or TuSimple. This is incredibly expensive, with little visibility of when they would ever turn a profit. Shares in US-listed Aurora and and TuSimple have dropped 80% and 70% respectively since the beginning of the year, while Argo AI laid off 150 staff in July and GM cancelled its IPO of Cruise back in March.
Roadside infrastructure would reduce their burden significantly. The bottleneck right now for independently operating self-driving cars are the vehicle sensors, which suffer blindspots. Roadside sensors would help cut those. UBS reckons China is planning to install, per kilometre, 50 cameras, 20 lidars, 20 millimetre-wave radars, 10 intelligent roadside units (with mobile connectivity), and four oxygen sensors on roads with selective autonomous driving support, with commercial adoption from around 2026.
This is where it gets tricky when looking at UK applications of a similar solution. China can pay for this by increasing road tolls, especially for commercial vehicles. We’re still debating tolls themselves, with every subsequent government kicking the debate about how we pay for road use in the future into the long grass. We’re also far more wary about the privacy issue, something that’s not up for debate for China’s heavily monitored citizens.
China’s centralised control can also mandate solutions that would entail lengthy agreements between private companies and the state in the UK and elsewhere in Europe. “Industry needs to come up with a standard way before anyone can invest in infrastructure,” said Shelley Cheshire, marketing lead for Zenzic, a government-funded organisation encouraging investment in autonomous technology in the UK.
The governmental body tasked with steering strategy in this area, the Centre for Connected and Autonomous Vehicles (CCAV), is putting the onus on the vehicle companies. “In the early days, we are expecting that automated vehicles won’t be dependent on safety-critical external infrastructure because it’s not there yet,” a source there said under condition of anonymity. “It’s a bit chicken and egg.”
That’s not to say the UK isn’t looking at infrastructure. For example, ServCity is a government-funded autonomous trial in London’s Greenwich in which six self-driving Nissan Leaf hatchbacks (manned with a back-up safety driver) drive around the busy suburb with support from a network of roadside cameras and sensors.
Meanwhile, UTAC’s Millbrook test site in Bedfordshire has a site-wide connected network that allows companies “to fail in private” when it comes to relying on external data to help autonomously controlled vehicles, as UTAC’s Stoker put it.
As it stands, the UK government earlier this year paused development of Britain’s smart motorway network after it came under criticism for failing to protect those in broken-down cars. It might be that, in this country at least, we might have to stick to the dictionary definition of autonomous and let vehicle companies go it alone as we look to drive hands free.
Source: Autocar
