Are Self-Driving Cars Safe?

< Back to blog

Once the outlandish fantasies of science fiction, autonomous vehicles in one form or another are slowly becoming more and more present on our roads. But what are self-driving cars? How do they work? And are they safe?

Safety of self-driving cars

As with all novel ideas and young industries, there are growing pains that could make people skeptical of their safety. Unfortunately, self-driving cars aren’t an exception and their have been fatal incidents resulting from the use of autonomous vehicles. However, it is the safety potential of self-driving cars that is so appealing.

Incidents

Accidents involving driverless cars aren’t common but they’re not rare either, with a handful of cases, some of which were fatal, occurring since 2016.

Pretty much all of the leading autonomous vehicle manufacturers have encountered problems throughout their development process and both Tesla and Uber have been involved in fatal accidents in the United States due to software issues.

However, the very few incidents for the millions of combined autonomous miles driven does prove promising.

Safety potential

With the UK government wanting fully driverless cars on the road by 2021, there must be a strong reason why and many point to the potential safety of self-driving cars as the answer. There were 160,597 road traffic accidents in the UK during 2018 and most of which were deemed to be the driver’s fault.

Autonomous vehicles can help alleviate this issue due to the fact that they are model drivers. They won’t speed, they won’t weave between lanes and, most importantly, they are constantly aware of their surroundings. However, there could be issues when both driverless cars and manually controlled vehicles share the road as one may do something the other can’t predict.

The trolley problem

As well as both technical and safety issues, driverless cars also have to deal with ethical problems such as the trolley problem. The trolley problem presents the situation in which a collision is inevitable and the vehicle must decide which collision scenario is most moral, usually choosing between the death of five or the death of one.

Ethical decisions like this are notoriously difficult and complex and some engineers have resulted in using public decision making in order to aid the machine learning process e.g. MIT’s Moral Machine. In contrast some manufacturers aren’t worrying about the ethics and ensuring that their drivers are prioritised at all times.

The six levels of vehicle autonomy

Experts at the Society of Automotive Engineers (SAE) have outlined six levels of autonomy for self-driving cars, that span from fully manual to fully autonomous, and it may surprise you to find out that your vehicle may already been deemed ‘Autonomous’ to some degree.

Level 0 – fully manual

Fully manual vehicles are currently the most common on our roads and require a human to manually control all aspects of driving. However, these are likely to become the thing of the past with the majority of new vehicles displaying at least some level of autonomy.

Level 1 – driver assistance (hands on)

Level 1 vehicles are a small step towards autonomy and have one automated aspect across steering, speed and braking. For example, a vehicle with automatic emergency braking or adaptive cruise control could be deemed as a level 1 vehicle.

Level 2 – partly automated driving (hands off)

Vehicles display partly automated driving when their systems take full control of accelerating, braking and steering. This can be seen in modern vehicles that offer smart parking or lane control features that can manoeuvre or park the car without driver input.

Level 3 – highly automated driving (eyes off)

It’s not till level 3 that we get a glimpse of true self-driving, where the driver isn’t required at all in certain scenarios e.g. motorway driving. However, the driver will still be required to take control at a moment’s notice.

Level 4 – fully automated driving (mind off)

Level 4 vehicles provide fully automated and true self-driving functionality, allowing the driver to sit back, relax and even take a kip if they wanted to. However, they still allow the driver to take control and require the driver to do so in areas where fully automated driving is currently prohibited.

Level 5 – full automation (steering wheel optional)

The highest level of automation and the end goal of self-driving vehicles is full automation, which requires no human intervention at all. With level 5 vehicles drivers don’t need to be fit to drive or even hold a license and many don’t include a cockpit or steering wheel, rendering everyone in the car a passenger of the vehicle.

Waymo Driverless Car
Grendelkhan / CC BY-SA (https://creativecommons.org/licenses/by-sa/4.0)

How do self-driving cars work?

Organisations have taken clever approaches to explain driverless vehicles in the past, notably Nissan‘s partnership with Lucasfilm in order to use Star Wars as a medium to explain their vehicles, but it’s not as complicated on paper as you may think.

As with all Artificial Intelligence systems, self-driving cars attempt to mimic and build upon the innate human ability to think and act rationally. They do this by sensing their environment and through decision making based on their surroundings and own knowledge.

There are three key components necessary for self-driving cars: radar, cameras and lidar.

1. Radar – object direction and speed

Radar is the underpinning of numerous safety tech like adaptive cruise control and automatic emergency braking, and provides reliable object speed detection regardless of weather.

However, radar fails to distinguish the objects themselves and can only identify their speed and direction. As such, radar alone wouldn’t be able to separate a car from a pedestrian or bicycle for example.

2. Cameras – object recognition

Self-driving cars utilise a multitude of cameras in order to provide the vehicle with sight in a manner similar to human vision. They can only see what’s illuminated and are affected by bad weather, but they have impressive resolution and provide great amounts of detail about the vehicle’s surroundings.

Cameras are used in conjunction with AI techniques, like machine learning, in order to provide the identification of objects that radar lacks. The vehicle can be taught to pick out speed signs, lane lines, pedestrians etc. allowing the vehicle to navigate and respond to threats.

3. Lidar – object location

The final piece of the autonomous vehicle puzzle is lidar, which builds a map of the vehicle’s surroundings. Lidar uses millions of infrared light pulses a second in order to survey the environment by measuring how long it takes to receive reflections. Differences in response times and wavelengths are then used to make 3-D representations.

While Lidar doesn’t provide the same resolution as cameras, it can be used in various lighting conditions and provides a sense of object location in the vehicle’s 3-D representation of its environment.

Example lidar point cloud
Daniel L. Lu / CC BY (https://creativecommons.org/licenses/by/4.0)

Examples of self-driving vehicles

While there aren’t any level 4 or 5 autonomous vehicles currently on the market, it’s likely there will be soon. Tesla CEO Elon Musk quoted 2020 as the year their level 5 Autopilot software will roll out and both Uber and Google are well into their own projects with ATG and Waymo respectively.

However, there are numerous level 2 vehicles on the market and some, like Tesla’s AutoPilot vehicles, that nearly fulfil the level 3 requirements. This means you can get your hands on a partly automated lease car easier than you think, with practically all manufacturers having developed their own assistive software such as Mercedes‘ Distronic, Nissan’s ProPilot and BMW’s Personal CoPilot.

Self-Driving vehicles aren’t restricted just to consumer driving either. Numerous companies are also using the technology for last mile delivery with the likes of Starship delivery robots already in use in the UK in Milton Keynes and the UK’s Kar-Go in the later stages of its development.

Regardless of current standing, self-driving cars are the future of the automotive industry and provide advancements in both accessibility and safety. Even though there have been issues in the past, developments in autonomous technology is likely to further solidify self-driving cars as safe and reliable vehicles.

For the latest car reviews and auto news, visit the Moneyshake blog to keep up to date.