Great progress has been made in the quest towards self-driving cars, but there is still a long way to go before truly autonomous vehicles can become a commercial reality. Let’s look into the various levels of car autonomy, and see what’s still needed for cars to really become self-driving.
Motoring enthusiasts may not enjoy hearing this, but the majority of drivers on our roads would really rather be doing something other than actually drive their cars. In reality, enthusiasts make up a very small percentage of motorists, while the remainder see driving as a sort of necessary evil – a chore to complete every time they need to get somewhere else.
It’s with these drivers in mind that car manufacturers and software engineers have been working on developing the technology needed for fully self-driving vehicles. The end goal would basically be for a traveler to get into the car, enter their desired destination into the car’s control system, and then leave the whole process of getting to that destination to the vehicle’s own devices.
A typical array of sensors, cameras and processors, needed by an autonomous car.
With this in mind, a host of technologies are already being applied to high-end cars. High-speed adaptive cruise control with lane-keeping assistance and automated braking are becoming commonplace, and various cameras and sensors feed the car’s control system with the data it needs to become aware of its immediate surroundings.
Levels of car autonomy
As with most goals in life, getting to the point where cars could drive themselves will take a number of steps, with each step adding extra capabilities to the car’s behaviour. With this in mind, the Society of Automotive Engineers (SAE) has classified various levels of autonomy according to a vehicle’s abilities and the amount of driver intervention required.
Level 0: While some car control systems might be automated, these systems can only momentarily intervene or warn the driver of hazards, but full control of the vehicle is still the driver’s responsibility. As an example, think of lane-departure warning, in which the assistance system alerts the driver but takes no action to correct the car’s trajectory.
Level 1: The vehicle’s autonomous systems share control with the driver, but the driver is still the ultimate decider about the car’s actions. This level of autonomy includes either steering assistance (as in self-parking or lane-keeping assistance), or adaptive cruise control (where the driver still has to steer, but the car regulates its own speed and brakes while the system is active).
Level 2: This is the next step towards autonomy, where the driver assistance systems control both the steering- and speed regulation when activated. It still requires the driver to remain attentive on the proceedings, and be ready to take full control and override the assistance systems when needed. Most current semi-autonomous cars operate at this level.
Level 3: For those who cannot resist texting and driving (hint: it’s always a bad idea!), level 3 will be the best news ever. The car will be able to handle most driving tasks by itself when the system is active, but can request intervention from the driver in response to unexpected hazards. This still means limited functionality, though, and driver inputs will still be required in many instances.
Level 4: This is much the same as Level 3, but with the additional ability for the car to pull over and park itself if the driver doesn’t resume control when the system encounters conditions which fall outside its abilities. Level 4 automation can handle a wider variety of conditions, but still needs a driver to take over if needed.
Level 5: This is where the driver becomes redundant, because the car can handle all conditions and hazards by itself. This holy grail of self-driving cars is still a dream, however, although real-world testing of level 5 autonomy has been going for years already.
What obstacles still lie ahead?
With the current state of technology, a semi-autonomous car will be aware of its immediate surroundings, and will even take evasive or protective action once it senses imminent danger. Failing that, it will hand control back to the driver, leaving life-and-death decisions to the human who is still nominally in control.
The main problem is that almost none of the cars currently on the road are really capable of driving themselves, leaving them prone to erratic and irrational actions from their drivers. This is problematic for a computer system which is programmed to respond in a logical and rational fashion, because it can’t accurately predict what the road- and traffic conditions will be a few seconds in advance.
To circumvent this, additional information is needed, and this information will need to come from the nearby infrastructure and surrounding cars. In short, a truly self-driving car will also need to be able to talk to all the cars around it, as well as to roads and buildings, to form a complete picture of its operational environment.
Once this type of network integration becomes feasible, self-driving cars will be able to safely navigate to their destination, relying on their own cameras and sensors to form an image of their immediate surroundings, while synchronising their route planning, speed and lane choices with the connected network of nearby cars and the available infrastructure.
5G may be the key
Managing this much data is sure to be a challenge, especially with cars also turning into connected devices, but vast reserves of data transfer capability and a standardised communication protocol will go a long way towards interconnecting cars with each other and with their environments.
5G shows much promise of being able to handle this huge data load, and with the Internet of Things (IoT) connecting cars and their surroundings, the self-driving algorithms embedded in future cars will be able to form complete images of their environment from various sources, ensuring operating efficiency and safety for their occupants alike.
While we’re already enjoying some of the technologies which will eventually contribute to full driving automation, some crucial elements are still in development at the moment. But, once everything is connected (even human-driven cars will be hooked into the network, integrating their location and speed into the control algorithms), self-driving cars will be able to function to their full abilities. Until then, we can take advantage those comfort-enhancing semi-autonomous functions, whether we enjoy driving or not.