With all of the excitement around Autonomous Vehicles, we’re looking at the history that has brought this enormously promising industry to where it is today and some of the challenges that may continue to make it a rocky road.
BY MIKA SKARP
At mere imagination stages back in the 50s, Autonomous Vehicle technology development has been moving like a pace car for decades.
The Real First Autonomous Cars
The first landmark tests were staged as a series of self-driving cars or autonomous vehicles derbies in the Mojave Desert in Nevada by DARPA, the Defense Advanced Research Projects Agency.
The stiff competition pitted 132 contenders against 2004 Grand Challenge winner Carnegie Mellon (an early innovator showcased on Jay Leno in 1995), with their then newest model, “Sandstorm” a rather crude looking red Humvee.
The next year saw Stanford taking the super-sized $2 Million prize “Stanley” a Volkwagen Toureg, but Carnegie Mellon came back fighting in 2007 with its Chevy Tahoe crowned “Boss.
At each of the DARPA events, (and several one off demonstrations in Europe and Asia), while the tactical, obstructional challenges increased, the speed of the vehicles, starting at a sleep-inducing 10-15 mph, shows how far we have come in just a decade.
Increasing performance of self-driving cars
Today’s CAVs (Connected Autonomous Vehicles) are achieving near flawless performance at well over 55, and even picking up speeding tickets along the way. In urban centers, spare some highly publicized, and in some cases tragic incidents, CAVs are proving the future of transportation.
“Connected Autonomous Vehicles (CAVs) are proving the future of transportation”
Still, this is just the beginning.
While autonomous driving was introduced a couple years ago by mostly high end car brands, their design was focused on making highway driving smoother and parking easier and more relaxed.
Today, from ride hailing services like Uber and Lyft to logistics trucks, Autonomous Vehicles are buzzing about the streets of nearly every major metropolis from LA to Paris to Shanghai.
And as compelling as that has become for the automotive industry, tech giants like Google (Waymo) and a blossoming ecosystem of smaller innovators developing improved cameras, sensors, LiDAR, map data compression and all manner of connectivity enhancements there remains some uncertainty about when the driverless experience will pull up the curb in earnest.
That said, there is no doubt that while costs for these CAV peripherals are coming down, in order to bring the Autonomous Car into range of the average consumer manufacturers need to reduce number of laser, radars and video cameras and trust mobile networks.
So the thinking goes, because Mobile network are a shared resource, they are much more cost effective than custom built hardware. Still the problem with mobile networks is that capacity and reliability are still uncertain, and this needs to end.
While we are fast approaching an Autonomous Vehicle that delivers on the promises of total safety and efficiency, several high profile fatal accidents have given some pause. Uber, ever eager to cut the chord with its costly pool of drivers even halted its program while the bugs get worked out.
This is an important point about the future of the industry as a whole but not one that would move it into the slow lane. Software issues can be fixed, but human error, which accounts for 93% of today’s road accidents, can’t be.
“No doubt CAVs will be the safest, smoothest, least polluting and most efficient way to drive”
No doubt CAVs will be the safest, smoothest, least polluting and most efficient way to drive, but there are some hurdles yet to jump.
Here in Finland, where autonomous vehicles have been on the road for some time, we are at the forefront of driverless cars negotiating extreme conditions, with VTT Technical Research Centre’s Martti, and his more urbane other half, Marylin, built for city driving.
But fully robotic, un-manned vehicles, or “land drones” like this capable of negotiating any street or highway raise serious concerns. It is not hard to imagine that they would quickly go from delivering passengers and packages to much more incendiary play loads as a weapon in the hands of terrorists.
It’s no wonder that these innovations bring equal amounts of enthusiasm and fear. While this may not be the primary driver (pardon the pun) of new legislation, like that enacted in California requiring CAVs to maintain a constant, failsafe mobile connection, it is but another very good reason.
While CAVs, with their array of cameras, sensors, LiDAR and GPS don’t need a mobile connection to move around safely and needn’t necessarily consume a lot of bandwidth, Barclay Analyst Brian Johnson suggests that with all of the data they collect and share could well consume up to 100 Gigabytes a minute. Interviewed for CNBC Johnson paints a startling picture.
Autonomous driving needs network capacity
“Assuming the entire US fleet of vehicles (260mn vehicles) has a similar data generation, it would create an ocean of data. To put it in context, one hour’s worth of raw data across the entire US fleet would be ~5,800 exabytes in size.”
Regardless, any truly unmanned or driverless car must be accessible by remote control, and, in the case of extreme situations, allow remote shut off.
The only way to achieve this is via the mobile CAV are satisfied, these connections must be at least very fast, provide ultra-low latency and 100% reliable.
And that connection, whether for a CAV or any logistical device needs to be there regardless of any other strains or congestion within the network.
While the issue of coverage is a different problem entirely, the essential point here, and as clearly outlined by the state of California, the vehicle must be connected at all times, or if a connection is lost the vehicle must be stopped remotely.
Network Slicing provides a perfect tool for autonomous vehicles of all kinds. At the high level, the use case looks like this. As soon as the car is started it will request a secure, SLA-assured connection from the network.
How does a self-driving car determine its route and assure network coverage?
Knowing its route in advance, the vehicle will communicate its route to the network to make capacity reservations in advance. Routes with capacity issues, whether due to congestion or outage may then be re-entered into the car’s mapping system such that it can suggest different routes that ensure flawless coverage.
There is no doubt but that elephant in the room is the lack of a complete cellular network.
Certainly the arrival of eSIMs that can allow the device, in this case, the car to “hop” from one network to another based on availability. This will go a long way toward satisfying legal requirements, but what happens between LA and Los Vegas won’t stay in Vegas.
The problem of (4G/5G) network coverage
Though this is less of an issue in smaller, superbly connected countries like Finland, in the US there are giant swaths of the country that don’t and may never get adequate 4G or 5G coverage.
Aside from the high cost of building out the networks in large remote regions, the business case for doing so is far from clear. But that all changes with the arrival of this new and most promising industry.
If selling slices to the automotive industry, transport logistics and individuals provide the business case, which it certainly seems to, wouldn’t building out 4G and 5G corridors everywhere our cars can go suddenly make a lot more sense.
Some other solutions are on the horizon as well, and ones that may even pose a competitive threat to carriers. Mesh networking is a great example.
This is the ability to use mobile devices, in this case, the vehicle itself, as a network repeater to allow users to leap from to the closest available base station via one or multiple bearers in between. This has already been put into place in remote regions of the developing world with companies like RightMesh.
Here, a new data sharing ecosystem is coming into play where users can re-sell their unused data to other users on the network. It’s not only a great way to extend coverage with a built-in incentive for users but may save carrier billions in rural network build-outs.
On the downside, it has the potential of eating directly into the carrier’s revenues at precisely a time when they need to shore them up.
The best scenario would be for Carriers to take a firm grasp on the wheel and begin testing network slices, (ideally eSIM-powered slices capable of jumping networks), and charge premiums drivers or the car manufacturers themselves.
Even if the CAV technology is some ways away from becoming mainstream, the business models are there, and they directly address some of the biggest regulatory hurdles this nascent industry faces.