The automotive industry is fully committed to developing self-driving technology. This fact is evidenced by the ongoing thrust of automotive exhibitions, as well as continuous innovation in the automotive industry.
The development of SAE level 4 vehicles — that is, vehicles that are fully-automated and able to perform all functions themselves — is ongoing. Self-driving car manufacturers are actually hoping that these vehicles will be available to the public within the next decade. So, what technological advancements have happened to make this a possibility? And what are the gaps in self-driving car abilities that must be closed before it can hit the mainstream?
1. High-Tech Sensors
Self-driving cars use hundreds of sensors to control speed and assess hazards. Some of these types of sensors already feature on the cars being driven today. For instance, you may already be enjoying the safety benefits that come with lane-changing assistance, blind-spot checking sensors, and parking sensors.
In a self-driving car, multiple sensors collect a vast amount of data, which is then processed in real time by the vehicle computer. These sensors need to replicate everything that a human pair of eyes does while driving. Human eyes send signals to the brain, which then assesses danger and allows people to navigate the road, to steer correctly and adjust the speed using the information currently available.
LIDAR
The primary system of sensors to be used by self-driving cars is LIDAR, which stands for light detection and ranging. LIDAR technology was first developed over 50 years ago as a 3D mapping tool. It has since been used extensively by archaeologists and to collect data during space missions.
LIDAR uses powerful pulses of light to measure distances from all objects in a 360⁰ radius by calculating how long it takes the reflected light to return to the sensor. Millions of pulses of light are dispatched per second, which means a self-driving car can maintain an incredibly accurate picture of where it is in relation to all other objects.
Because it is mounted on the top of the car, LIDAR can maintain (unlike the human eye) continuous sight in all directions. It can also calculate distances within a tiny margin or error — around two centimeters — which is an improvement on human abilities. It can even produce data that will enable the car computer to calculate the trajectory of moving objects.
To supplement LIDAR, a self-driving car will also use information from lower-positioned radars to make assessments in bumper-to-bumper traffic and to make parking calculations.
2. 5G Connectivity
The existing 4G networks have been fast enough to be capable of supporting online gaming and video streaming. However, it is the introduction of new 5G networks that will make self-driving cars a possibility for the mainstream.
To ensure vehicle safety and to improve public confidence, it is vital that self-driving cars be capable of making decisions as quickly as humans can, or even faster. The average human decision-making process when it comes to assessing danger and beginning to act on it is just a few milliseconds.
It is expected that 5G will routinely be able to match and even improve upon this speed. Fifth generation technology will connect multiple technologies with an ultra-fast, highly responsive and reliable network.
A self-driving car’s sensors collect a vast amount of data that needs an incredibly fast network to be processed in real time. Connectivity via 5G offers this capability, and once it is widely adopted, it is expected that it will reach speeds 100 times faster than 4G. Thus, through the use of 5G technology, it is fully anticipated that self-driving cars will then be able to act faster than the average human reflex.
Moreover, 5G connectivity won’t just allow the car sensors and computer to communicate internally. Vehicle-to-vehicle (V2V) data sharing will mean that vehicles will also be able to alert each other and, thus, adjust for unexpected changes to driving conditions, such as road debris or weather fluctuations.
3. Sophisticated Machine-Learning Algorithms
High-speed connections bring a vast amount of data to a self-driving car’s computer every second. How does it, then, use this data to ensure the car continues to move safely, avoids collisions, and reaches its final destination as quickly as possible? This is the role of machine-learning software.
To successfully maneuver the car, the software must replace the human driver’s brain in object detection, object classification, object localization, and movement prediction. To do this, it has to operate using machine-learning algorithms, a subset of AI.
Algorithms use comprehensive, strictly set processes to make calculations and problem solve. In self-driving cars, the types of complex algorithms used can make predictions with a high degree of confidence and derive value from vast data sets by detecting patterns and anomalies.
These algorithm computations happen incredibly quickly, meaning a self-driving car is always ‘alert.’ Unlike a human driver, the use of these algorithms is not impeded by tiredness, mood, or in-vehicle distractions.
However, these systems must be rigorously tested to ensure public safety. Researchers are continuing to not only develop sophisticated algorithms but also ways to comprehensively test them.
Conclusion
It is the advancements made in these three areas that have led to the point of level 4 self-driving cars being almost ready for mainstream use.
However, before anyone ever gets to see them for general sale, honing the required technologies even further, and comprehensive, rigorous testing are the main tasks for developers of self-driving technology. These are among the many interesting talking points at self-driving technology conferences as the industry moves ever closer to realizing its goals.
Author Ahmed Bahrozyan is the Chief Executive Officer of the Public Transport Agency - Roads and Transport Authority in the United Arab Emirates which is responsible for providing for the needs of public transport in the city. He is also Chairperson of the Dubai World Congress for Self-Driving Transport organizing committee.
