As connected vehicles become mainstream, we’ve been bombarded with headlines suggesting that driverless vehicles are nearly upon us.
Although driverless vehicles are currently being tested in various locations around the world—see “Driverless car tested in public in UK” or “Otto driverless truck makes 120-mile beer run in Colorado”—quite a few significant challenges lie in the path of widespread deployment such as technology, the business case, and cultural factors.
Here’s a scenario for the next few years (as I imagine it), based on my work in the field of future wireless communications that will be crucial to how all of this unfolds.
Connected vehicles will become the norm because vehicle-to-infrastructure (including various elements of a vehicle’s immediate environment, as well as the cloud) is rapidly becoming commonplace. That connectivity, driven by safety and infotainment applications, is already integrated into automakers’ current manufacturing and future plans.
Driverless vehicles, largely for relatively low-speed, metropolitan deliveries, may also appear (perhaps akin to golf carts on steroids) already well-covered by existing cellular infrastructure. Still, a brave human may need to be onboard for making actual deliveries. It’s also possible driverless transport on major roadways between metro areas will take hold. I anticipate an incremental approach. Technical, business, practical, and cultural factors must line up to make this happen.
However, the advent of an all-driverless world runs into a number of real-world constraints. Many people enjoy driving and there are currently some 1.2 billion vehicles on the road worldwide. So, let’s look at the drivers for connected and autonomous vehicles, the technical communication, processing challenges, and business case that together, may block our road to a driverless future.
Safety, Cost, In The Driver’s Seat
It’s important to recall that safety and cost considerations have propelled the trend in connected cars. From roadside assistance programs like OnStar to policy directives from the U.S. Department of Transportation, the theory is that connectedness can alert drivers to impending collisions and notify them of imminent mechanical failures. As vehicles become more electrified and their costs get more tied to their electronics than to mechanical overhead, their weight drops, while operations and maintenance costs decrease accordingly.
Safety is also cited as a driver for autonomous vehicles, but there are lessons for the experience of railways, where public perception plays an important role. The constrained rail environment makes autonomous operation easier, and indeed this year marks the fiftieth anniversary of the opening of the first automatic heavy rail line- the Victoria Line in London. However, then, as now, a driver sits in the cab for emergencies and operating the doors.
There are now about 1,000 km of completely driverless metro lines worldwide, a tiny fraction of the 1,300,000 miles of conventional railways. Indeed, the Federal Railroad Administration in the U.S. cited safety reasons behind a recently proposed a minimum of two crew members on all trains.
Sweet Spots for 5G
Connectedness for safety and cost considerations is where mobile telecom advancements will really matter. This is particularly true for driverless vehicles.
Currently, telecom carriers are still reaping a return on investment from their 4G networks, while 5G remains in development with numerous technical challenges remaining. In fact, “5G” is a bit of a misnomer because its advancements will look nothing like today’s 4G capabilities. Once realized, 5G will usher in decades of innovation.
Without getting mired in technical details, 5G technology will bring significantly higher data speeds, lower latency, a level of network flexibility, and other features that can be applied to enhance safety for vehicles with drivers. 5G will allow multiple network types to share the same equipment and support inter-vehicular communication, where each vehicle would represent a network node. Inter-vehicle sensing will help prevent collisions and an intra-vehicle network will troubleshoot and predict mechanical failures, while communicating its findings with the cloud.
Challenges Remain
Even if 5G fulfills its promise, it may not overcome certain constraints—some technical, some business model-related, in the connected and driverless vehicle models.
If connected vehicles rely on external communications infrastructure such as a cellular network, they’ll need connectivity mostly in congested urban environments where collisions are more likely. Having said that, too many simultaneous participants in such a network will hinder low latency and effective vehicle-to-vehicle communications needed for collision avoidance.
Conversely, if we consider a vehicle-to-infrastructure model for driverless vehicles in large swaths of suburban or rural roadways, the cost of external, infrastructure-based sensors and two-way data transmission becomes prohibitively expensive.
Autonomous Vehicles = Greater Challenge
Some argue that driverless vehicles will need to be truly autonomous, which means they operate without constant data contact with their external environment. Instead, this type of vehicle would need to sense its environment and use artificial intelligence (AI) to make critical judgments in real time.
First, the vehicle must know exactly where it is, and have very good awareness of its surroundings. The use of cameras to provide that information is being explored, though that will require incredibly fast processing to be useful in near real time.
Second, the AI needs to make judgments to successfully pilot the vehicle, which presents a daunting challenge. Humans are really good at that sort of complex information management and decision-making. However, we’re still a long way from getting AI to compete with us when lives are at stake.
Mapping the Way
Another option being explored would have the vehicle work within a known environment with an accurate map of its surroundings, which poses its own challenges. (See the article, “Building a Road Map for the Self-driving Car,” from The New York Times’ 3 March 2017 edition.)
Such a vehicle would know the location of the road, possible dangers, traffic delays, and be on the lookout for anything that might cause a problem. As it detected something new, out of place, or unexpected, it would have to react and adapt in near real time and likely communicate with the cloud, which would provide other autonomous vehicles with updates to their internal maps. Since every autonomous vehicle would be engaged in this fashion, that’s a lot of data being transmitted between the vehicles and the cloud, again, potentially challenging a network’s capacity.
Such environmental data might be transmitted between nearby vehicles on a point-to-point basis and would include the driving intention of each vehicle. If one car slows to avoid a sudden obstacle, there’s clear value in transmitting that information to other nearby vehicles, perhaps in milliseconds. In an urban environment with congested traffic that’s not moving rapidly, extremely low latency of wireless signal may not be needed, but extremely high reliability would be required.
Finally, consider the implications for insurance. Today you have a driver, vehicle, and set of rules to govern responsibility. The driver or (in the case of mechanical trouble) the car, is responsible in a collision. Both can be insured. In the chain of events described in the foregoing examples, where does fault lie in a collision? Wireless network operators are unlikely to broaden their exposure if it’s proven the network had a glitch just prior to impact.
Connected Is Safer, But…
It’s clear (at least to me) that connected vehicles can succeed and enhance safety. Driverless vehicles will have limited applications until (or unless) we reach a very high level of confidence that the daunting technical challenges that lie ahead have been resolved to the degree that we’d bet our lives on it.