Full Self-Driving Cars Are Still A Long Way Off – Here’s Why

A few days ago, the UK Department of Transport announced a consultation on Automated Lane Keeping Systems (ALKS). This led to media outlets such as the BBC arguing that hands-free driving might be legal in the UK as soon as Spring 2021. But some went further, claiming this was a step towards self-driving cars by next year. Automated safety features are a significant part of cutting-edge car technology, so a major feature in most electric vehicles, although far from exclusively for EVs. Tesla
TSLA
in particular makes a lot of noise about its self-driving abilities, and contender Lucid is claiming it will go further with its Air. But is this really all hype, and will we ever really be able to sit back and let the cars do all the work?

The first thing to clear up is that what the UK government is talking about is not even close to full self-driving. The consultation is regarding Automated Lane Keeping Systems (ALKS), which are essentially a more advanced version of cruise control that can operate steering as well as speed. As roads get ever-more congested and dotted with average speed limit sections, cruise control is becoming an essential feature to ease your pain in traffic. A basic system simply adjusts your speed to maintain a specific mph. The more advanced adaptive versions like Mercedes’s DISTRONIC PLUS use sensors to detect the vehicle in front and will apply brakes automatically if you get closer than a predefined distance. This is safer as well as more relaxing.

ALKS goes one step further and will actually steer your vehicle for you. Lots of cars now feature Lane Keep Assist, but this merely detects road markings on highways and will notify you if you stray towards the edge. Sometimes this is just an audible reminder, but Vauxhall’s new electric Corsa-e, for example, actually gives the steering wheel a little tug to make sure you noticed. With ALKS, however, the car will use its sensing of adjacent vehicles and road markings to corner for you, so you have to do even less to control the vehicle. Tesla’s Autopilot already includes this ability, although you are supposed to keep your hands on the wheel at all time. One man in Hertfordshire was banned from driving for abusing the function, and there have been a number of crashes in the US due to people eating or drinking while Autopilot was enabled. One Tesla driver even posted a video of himself eating a burger with his hands nowhere near the wheel.

It’s worth mentioning at this point that there are five levels of driving automation, or six if you include Level 0, which means the car has no automation at all. Level 1 includes the adaptive cruise control and Lane Keep Assist functions already mentioned. Interestingly, Euro NCAP has been testing Autonomous Emergency Braking (AEB) since 2014, and it is now impossible to receive a five-star rating without this feature. Level 2 is described as “Partial Automation”, and this is where ALKS sits. The driver still must keep their hands on the wheel, ready to take control if necessary, but the car can control speed and steering, even around bends. Apart from Tesla’s Autopilot, Volvo Pilot Assist and Audi Traffic Jam Assist fit into this category.

Level 3 is where we start to get towards self-driving. This will only be under ideal conditions like highways and at certain speeds. The driver still needs to be ready to take control if necessary but doesn’t always need their hands on the wheel. Level 4 is close to full autonomy. You can enter your address and the car will drive to it unaided. This has been demonstrated under test conditions, but regulations prevent it from being released yet. Level 5 is like Level 4 but with no human interaction necessary, so there is no need for a steering wheel or pedals in the car.

Elon Musk has already claimed his Tesla cars will have Level 5 self-driving in 2020, and all Teslas are bristling with sensors pointing in every direction ready for this. However, the only feature currently available that goes beyond Level 2 is Smart Summon, where you can get your car to come to you from a parking space. But Tesla owners haven’t found Smart Summon to be very reliable just yet, and it’s not legal in Europe so hasn’t been released. Waymo is a company that was spun off from Google
GOOGL
, and is now a separate subsidiary of Alphabet Inc
GOOGL
, Google’s parent company. Waymo has clocked up 10 billion miles of simulated autonomous driving and 20 million in the real world including with its customized Jaguar I-Pace test cars equipped with 5^th Generation Waymo Driver (5GWD), which is a trial of Level 5 abilities. There are a number of similar trials, like GM Cruise, but all still at the testing stage.

As the UK government consultation argues, ALKS and other autonomous technology has the potential to improve road safety, even if for car purchasers the main benefit is stress relief in traffic. There were 1,870 reported road deaths in the UK in 2019, and an incredible 38,800 estimated casualties to car crashes in the USA. Anything to help reduce this can only be welcomed. But the leap from this Level 2 autonomous feature to Level 3 and beyond is a much bigger one than the mere numerical increments would imply.

ALKS, like less sophisticated forms of cruise control, simply maintains the course a human being has chosen. It won’t change lanes, merely keep you in the same lane, steer along with road dynamics, and try to maintain your chosen distance to the car in front. In an emergency, the human driver is supposed to step in to make the difficult choices. Level 3 and above, however, control the car much more proactively. With this proactivity comes decision-making, responsibility, and (dare I say it) ethics. That’s a considerable change and is why Level 3 and above will take a much longer time to arrive than 2021.

The issue is not a technical one, or at least not predominately. Although a few cases have been widely publicized, Waymo self-driving cars have been involved in very few crashes during their 20 million miles on the road – just 11 in California in 2018, for example. Interestingly, a lot of these were rear-end collisions. The evidence seems to imply that the autonomous cars did something unexpected that caused a human driver to hit them, specifically braking unexpectedly. This kind of thing already happens with Autonomous Emergency Braking, which can react many times more quickly than most human beings and will slam the brakes on without regard to how close you’re following behind.

The wider issue here is related to one about artificial intelligence in general. The word “artificial” is there for a reason. An autonomous vehicle is not a human being, and as much as it may have been trained and programmed, it still won’t drive entirely like one. The classic cliché here is what it might do when faced with a child in the road – kill the child, or crash the car killing the occupant? Above that is the question of who is responsible for any deaths caused by an autonomous vehicle. The human owner/driver, the vendor of the autonomous system, or somebody else?

These are extremely difficult legal and ethical questions. They will take much longer to work out than the technology will need to function. In the first 2004 DARPA Grand Challenge, none of the autonomous vehicles entered managed to navigate a featureless 150-mile desert route successfully. In 2005, five did, and now we have vehicles driving around cities with very little incident. Technically, we’re very close to fully autonomous cars, if not there already. But society, law and morals aren’t ready, and won’t be for many years. While safety enhancements like ALKS will continue to filter through, fully autonomous driving that takes us from A to B while we sleep is still a long way off. Possibly decades.