The idea of self-driving cars flooding the roads is still a way off yet from becoming a reality. And understanding where the industry is in the pursuit of this can be quite confusing. Many companies have entered the autonomous space, but it seems driverless vehicles still haven’t pushed through the beta stage just yet. 

With wildly differing estimates of how many years we still have to go until the arrival of self- driving cars, we wanted to look at the main challenges facing full autonomy…

Safety considerations

Of course safety is a huge challenge when it comes to self-driving vehicles. Building an autonomous vehicle that can perceive the road better than a highly skilled human driver is difficult. The biggest advantage is that unlike some human drivers, a programmed vehicle will always obey the rules of the road, won’t speed, won’t get stressed, and won’t become distracted by a text message flickering onto a phone. 

Hypothetically, it’s believed that autonomous vehicles could detect things quicker than a person, especially at night or in low-light conditions, and react faster to avoid a collision. For example, many autonomous vehicles can use a combination of sensors and software to build a clear picture of the road. Light-detection, ranging and radar sensors work to map a real-time, 3D image of the environment around the vehicle, and can measure the size and speed of nearby moving objects. 

The current issues with this includes the fact that radar, while great at looking through heavy rain and fog, can’t make out complex shapes. And the light-detection and ranging sensors are short ranged and can be negatively affected by bad weather. When the environment proves too difficult and the weather too harsh, a person still must grab the wheel. There’s also the question of how the vehicles will be able to detect and avoid vulnerable road users, like cyclists, who can move more erratically. This real-world testing will is required, along with the simulations. 

Decision making 

Even when an autonomous car is able to fully perceive its surroundings, it must then be able to make key decisions. And every single decision it makes must also be validated, in order to ensure any safety concerns are being met. An autonomous car must be able to decide how fast to drive and when to change lanes etc. 

However, many factors can influence an autonomous car’s decisions; such as traffic, weather and pedestrian conditions, along with information that’s being sent by other autonomous vehicles and smart infrastructure in the area. Therefore, it’s essential that autonomous decision making is tested, redesigned and validated over billions of possible driving scenarios. Unfortunately, it’s impossible to so these tests in the real world within a reasonable time. 

Also, when an autonomous car makes a decision, it will need to strike a very fine balance between safety and usefulness. For example, not driving at all or driving very slowly will be very safe, but not very useful on the roads. It will fall to regulators to formalise the bounds of reasonable decision-making, which will allow automakers to programme their cars to act only within those bounds. 

The affordability issue

Another challenge facing automakers is being able to create a cost-effective autonomous car, that will persuade consumers to make the switch to driverless vehicles. At the moment, the technology required is still costing tens of thousands, meaning only a small percentage of businesses will be able to financially sustain rolling these cars out.

Individual drivers will likely not be willing to pay a high premium to have the new technology. Instead, automakers need to be able to build more precise systems than those that currently exist today, and they need to be made at a fraction of the cost.