‘When’ and ‘how’ now the biggest questions facing driverless vehicles

Autonomous cars are a big part of the future of transportation; there is no longer any serious doubt about that. With so many companies – from large and famous, to small and hungry – labouring to provide products and solutions for hands-free driving, not to mention the vast sums of money potentially to be made, a driverless future is inevitable.

Because there’s no question of if, the big unknowns regarding full passenger vehicle autonomy are when and how.

At first glance, it might seem as if we’re just on the cusp of the Autonomous Era. Examples abound: Google’s cute self-driving cars do a fine job shuttling passengers on their little journeys, while Audi’s driverless prototype last year transported a gaggle of journalists on a 600-mile or so journey from San Francisco to Las Vegas. Every incumbent automaker has at least one working prototype and their pilot programmes have generally been successful thus far. Advanced Driver Assistance Systems (ADAS) have moved quickly out of the realm of fantasy and speculation and will increasingly become part of our driving lives.

But there are plenty of roadblocks on the way to full autonomy. Firstly, getting a car to drive itself is a challenge that only deepens and intensifies the closer we get to complete hands-off operation. Also, a set of accepted protocols and standards for self-piloting vehicles is needed so that they can “talk” to each other, in order to avoid collisions and other mishaps. Lastly, and perhaps most challenging of all, is the almost complete lack of a regulatory framework for autonomous driving. In fact, one recent institutional development in an influential Californian market has seemed to throw up a big roadblock for self-piloting cars.

The push towards complete, hands-off driving has been rapid so far but we’ve many miles to go. This paper will assess the current state of the technology and its implementation, then detail the challenges – technical and otherwise – that will need to be surmounted in order for us to realise this most exciting transportation technology of the future.

Where we are now: gathering speed
In spite of the great brainpower, piles of capital, and good old-fashioned effort being devoted to autonomous vehicles, humans are still performing nearly all key driving tasks manually. If we grade our level of progress on a scale from 0 (driver only) to 5 (full autonomy), we’re currently less than halfway to the finish at level 2. Before we delve into the reasons why (and how we might progress further) let’s have a look at these five stages of automation, as defined by the International Organization of Motor Vehicle Manufacturers:

At the moment, we’re seated firmly in level 2, Partial Automation – ‘Driver must monitor the dynamic driving task and the driving environment at all times.’ The words in bold emphasise the fact that the person behind the wheel is essentially responsible for all major tasks and situational awareness involved in piloting their craft.

In other words, given the many elements involved in driving, comparatively little of it is automated. On a relatively clear motorway we can engage cruise control, for example, but we’re still responsible for steering, and have to stay alert to avoid obstacles and other vehicles ahead on our path. On busier routes, owners of certain recent models (like BMW’s offerings that use its ConnectedDrive technology) can take advantage of traffic jam assistant, which allows the driver to ‘glide’ along with traffic… momentarily freeing him or her to, say, dream of negotiating roads that aren’t stuffed with other cars. And, once we exit that motorway, parking assist features in certain recent models can help us not only manoeuvre our ride into a decent spot, they can search for, identify, and self-park in the space.

But in addition to requiring at least a little human input to engage those features, they’re quite limited in terms of both usefulness and function. After all, the bulk of vehicle operation takes place on surface-level streets with at least some semblance of traffic flow. Thus, our use of cruise control, traffic jam and/or parking assist is relatively infrequent and, as the names of the latter two imply, only on a lending-a-hand-basis to a driver that must remain engaged and alert during the process.

Much is involved in moving us from level 2 to level 5. Advancing up those three stages involves considerably more than a few technological and regulatory hops. The following section explores the many challenges, some rather daunting, that will slow progress.

The road to autonomy: machine, meet human
On each baby step on the way to full autonomy, the human/machine interaction (HMI) involved in driving a vehicle will have to be as smooth and seamless as possible. Thus far, the quite limited auto-features on cars have integrated well into human operation – to cite a classic example, cruise control requires only the push of a button or the moving of a lever to engage or disengage, for example.

As automation grows more complex and covers an increasingly complex set of tasks, solutions providers will have a more challenging time delineating human from automated jobs. Karl Iagnemma, CEO of nuTonomy, a young company that develops software and algorithms for autonomous vehicles, says that we’re still some way from perfecting that mix. According to him: “For systems that attempt to share control between the vehicle and driver (either by operating in parallel, or requiring a hand off), there is a lack of understanding of how to design safe, stable shared control systems. The HMI must evolve to a state where the human driver retains confidence in the vehicle's operation and also allows the driver to safely assume control of the system when required. To achieve this, future HMI systems must transmit the autonomous system's ‘mental model’ of the environment to the driver, in a way that is simple and easily understandable.”

Additionally, since we haven’t yet motored too far down the road to autonomy, there is the looming problem of standards and protocols. A dizzying number of early-stage companies are crowding the telematics and communications space, each probably hoping that their solutions will become the foundation for networked ADAS systems. Compounding that is the very active involvement of tech incumbents, which not only have their own ambitions in this regard but have also managed to become go-to names in certain other areas of gadgetry. Consider the success Apple has had in consumer electronics, with its i-line of consumer electronics, all of which are plugged into the company’s software and services ecosystem (iPhone users typically update their devices and obtain content through iTunes). Or Google, which has carved out a similar position with its Android operating system and Google Play store. Both companies have ambitions in the ADAS space, with platforms (Apple CarPlay, Android Auto) they’re trying to scale up in the likely hope that they’ll dominate. Google already has the big hardware, with its self-driving car efforts; Apple is strongly rumoured to be pursuing car manufacture on its own.

What’s limiting these potential players, large and small, is the acute lack of a regulatory framework for ADAS. Without this, it’s hard to know what functionalities and features to build into an ADAS system. Our current traffic laws work just fine for cars that are almost completely human-controlled. They will, however, become increasingly irrelevant the more automated our journeys become. In the European Union starting in 2018, all cars must be equipped with the eCall emergency contact system, which automatically notifies the nearest emergency centre in the case of a crash or other catastrophe. So solutions providers aiming to get a piece of the EU’s connected car market now know they will have to include eCall connectivity into their designs.

That’s a fine start but only a start. There is a raft of other laws that will have to be established in order to give a clearer picture of what “basics” an in-car system will require. These regulations cover a wide range of legal areas besides traffic law – also to be taken into account are telecoms regulations, rules about privacy, data protection statutes, to name but a few.

The challenge deepens as we get closer to cars that drive themselves. In December 2015, California’s Department of Motor Vehicles (DMV) drafted a set of regulations in anticipation of a self-piloting world, among which is a requirement that such craft be occupied by a person who holds a (non-existent, as yet) autonomous vehicle operator certificate issued by DMV. This individual ‘will be responsible for monitoring the safe operation of the vehicle at all times, and must be capable of taking over immediate control in the event of an autonomous technology failure or other emergency.’

The draft was scheduled to be submitted for public comment, and possible modification, in early 2016, in advance of becoming law. Regardless of what form it finally takes, it shows the great caution the authorities in the car-loving Golden State are showing in regards to autonomous vehicles. We can safely assume the ‘autonomous vehicle operator’ requirement will fall by the roadside once self-driving technology advances to the point where it is sufficiently advanced to ensure an exemplary degree of safety. But given the caution, it’ll likely be quite a few years before that time comes.

Autonomy anticipation: coming soon to an auto near you
While these issues are being addressed, the innovation caravan will continue to barrel down the road. In the immediate future, we should expect a near-constant evolution of existing ADAS features.

There are too many examples to list here. For one, BMW not long ago introduced a gesture system to its 7 Series, by which drivers can control certain functions with mere waves, rotations, etc. of their hand. Safety will continue to evolve – staying with those BMWs, new models have a set of practical, embedded safety features, such as one that corrects a car’s trajectory if it drifts out of lane. And we’ll see the connected car network more with the outside world – drivers of new Mercedes can access their Nest home heating systems while on the road, setting the temperature in advance for a toasty warm abode at the end of their journey. We can expect further developments in the regulatory sphere, too; in the US, the federal government’s Department of Transportation has promised to draft a uniform vehicle-to-vehicle communications standard by the beginning of 2017.

Again, we’re not on the brink of full autonomy yet. But that’s not going to stop solutions providers and carmakers from getting as close as possible to the ideal. Tesla’s recently-introduced AutoPilot feature set isn’t quite the machine-drive-thyself bit of magic its name implies but it does provide very helpful assistance like hands-free driving on well-marked, limited access motorways.

So, of course, the giant question looming over all of this is: when can we expect our cars to do all of the driving for us? Expert opinions differ, often wildly. In a recent analysis, investment bank Morgan Stanley pinned a date of 2026 when driverless vehicles will surmount the challenges in front of them and achieve full autonomy with no human control. But specialty research concern Victoria Transport Policy Institute is more pessimistic, saying that the Age of Autonomy will only really kick in “when autonomous vehicles become common and affordable, probably in the 2040s through 2060s, and some benefits may require prohibiting human-driven vehicles on certain roadways, which could take even longer.” And Daimler’s head of driver assistance and chassis systems Ralf Herrtwich anticipates that while cars will be able to function on their own in basic traffic and parking situations within a few years, “it will take at least another ten years before drivers won’t have to be involved in a vehicle’s operation at high speeds or in complex situations.”

So, yes, autonomous driving is the inevitable future of transportation. But let’s buckle up and get comfortable, because it’s almost certainly going to be a very long ride to get there.

Autonomous Vehicles