Autonomous driving: opportunities and uncertainties intertwined

There are a lot of ideas, heated discussions, and strong opinions about creating the city of the future. One important aspect of this vision is self-driving cars. This technology will affect every aspect of our lives. The way we travel, the way we live, creating new industries and jobs while disrupting other areas. The world is changing rapidly and we need to innovate to adapt to this new way of life. Car ownership in large cities is declining, and residents are finding it more convenient to hail taxi services. Malls and department stores are fighting for survival as online shopping increases. Busy lifestyles are becoming increasingly dependent on food delivery services. With the increase in online sales, innovative delivery services and package delivery are becoming mainstream technology trends. There is no doubt that it will change our lives now. Now is the time to look beyond the potential and see where we are in the process of implementing true autonomous driving; and how to move autonomous driving technology from "data collection mode" to large-scale deployment for daily life.

Let’s look at how far we’ve come in the evolution of self-driving technology, starting with the first Waymo car. Let’s look at California (the only state that is mandated by the DMV to disclose reports on self-driving vehicles tested). 28 companies reported results. Collectively, these companies drove over 2 million miles in self-driving mode and experienced 73,550 disengagements (i.e., the system handed control over to the safety driver, or the safety driver had to intervene to keep the car moving safely). This averaged out to 5 disengagements per 1,000 miles driven, or 1 disengagement per 200 miles driven. With this consistency, the California DMV relaxed its rules so that cars driving in self-driving mode will not have a safety driver on board. This is a significant shift, and the only company to have been granted this permit to date is Waymo. California also recently allowed self-driving for light trucks, which includes minivans, pickup trucks, utility vans, and step vans. This visibility in reporting provides a huge advantage in educating citizens; more details are needed to draw meaningful conclusions. For example, if a company reports a disengagement that occurred due to a perception issue, current DMV rules do not force companies to report the identification or classification of the object that failed (or repeatedly failed) the disengagement. In the absence of a standard process or method to verify the accuracy of the data, the effectiveness of reporting in improving public trust in technology is questionable.

Across the Pacific, China will become the largest market for autonomous vehicles. McKinsey's report said, "In 2040, autonomous vehicles may account for 66% of passenger miles, with travel services generating $1.1 trillion in market revenue and autonomous vehicle sales generating $0.9 trillion in market revenue."

In December 2017, Beijing was the first to introduce policies for autonomous vehicles on urban roads, followed by Shanghai. Subsequently, Beijing opened its first autonomous vehicle testing facility in February 2018. The testing facility covers an area of ​​133,000 square meters. These interesting ground-level test environments simulate real life and go through a large number of static and dynamic traffic scenarios on urban and rural roads. The Chinese government has invested a lot of money and resources to encourage companies to develop the most advanced technologies to grow rapidly in this market. China hopes to position itself as a technology leader in the autonomous vehicle market.

The European Union (EU) has clearly affirmed the universality of the benefits and impact of autonomous vehicles. Strong cooperation among countries to ensure widespread deployment with consistency and commonality has encouraged the large-scale deployment of autonomous driving technology in different countries. This is evident in the Vienna Convention on Road Traffic, which requires consistent regulation of cross-border traffic. It further strengthens the consistency of building highways connecting cities, making them conducive to autonomous driving testing. In addition, the EU has launched a plan to allow driverless trucks to operate between cities within two years. Reduced transportation time and fuel expenses are among the many expected benefits. Let's take a look at some countries in the EU. Germany, often considered the birthplace of traditional car companies, was one of the first countries to legalize autonomous vehicles on the road, but with a safety driver. The Netherlands has so far led not only in Europe but also in the world, and now allows the testing of autonomous vehicles without a driver and control. Sweden has provided a platform for testing small buses, and multiple companies have widely and ambitiously launched autonomous buses. "Soon, owning a non-autonomous car will be like owning a horse," said Carlos Moedas, the EU Commissioner for Research, Science and Innovation. With all these countries working under a common framework, Europe is building a safer and more efficient road network by developing a national strategy that incorporates automated driving.

Sensor technology is a key component in enabling self-driving vehicles. Sensors are able to detect objects in the vehicle’s path and act as the eyes of the self-driving vehicle. Let’s look at some of the sensors at the heart of self-driving technology.

Cameras are masters of classification and texture interpretation. Cameras today are the most commercially viable sensor choice. CMOS image sensors are the workhorse of cameras and have come a long way to take beautiful photos and selfies. These sensors have evolved from vision-only sensors to observation + perception. They are the primary sensors used for features such as Advanced Driver Assistance Systems (ADAS), traffic light recognition, and more. This sensor data delivers important information to the car’s decision-making process. Today’s CMOS image sensors for autonomous vehicles are highly advanced and can efficiently handle all driving scenarios, including some unique corners.

High dynamic range is a key parameter. This is the ability of the sensor to see the darkest and brightest parts in the same frame; without over-saturation or under-saturation. In other words, the sensor must not be affected by changes in ambient light, such as sudden exposure to bright sunlight or reflections, or entering or exiting a tunnel in daylight. Today, companies like ON Semiconductor are unique in offering image sensors with the highest dynamic range (HDR). Another trend in autonomous driving is higher resolution sensors. This is driven by the need for self-driving cars to see farther while having the ability to detect objects of various sizes. Essentially, it increases the ability to have more pixels per degree for objects, making the autonomous driving algorithms more efficient at object detection and classification.

Level 3 and higher autonomous vehicles rely heavily on hardware and software to make decisions, otherwise human decisions are needed. This also makes them more vulnerable to attacks. Therefore, features such as cybersecurity are added at the component level, that is, sensors. ON Semiconductor is launching the first automotive-grade image sensor with cybersecurity features.

LiDAR companies have also received a lot of attention from investors. This technology is considered very desirable for self-driving cars, but comes with a very high price tag. LiDAR's ability to create a 3D point cloud sets it apart from other perception technologies. Waymo has changed the playing field by introducing an in-built LiDAR that not only detects pedestrians, but also assesses which direction the pedestrian is facing and which direction they will move. This also enables the car to see hand gestures made by pedestrians or cyclists and adjust driving behavior accordingly. Range, resolution, and rotation/frame rate are other key parameters for choosing the type of LiDAR. Today's LiDAR can see up to 200 meters away. The concept is not only to detect objects on the road, but also to have the ability to classify them, provide feedback to the decision-making system, and change the course as needed. The reflectivity of the object is also crucial. While today's LiDAR is able to detect objects with high reflectivity at long distances, work is still underway to achieve long-range detection of very low reflectivity.

Radars, another commonly discussed sensor, are key to the future of autonomous driving. They are unique in that they operate in darkness and all weather conditions. Millimeter wave (MmWave) is a special type of radar technology that uses short-wave electromagnetic waves and is beginning to be widely used in autonomous driving.

By capturing reflected signals, radar systems can determine the distance, speed and angle of an object. Millimeter wave systems, operating at 76-81 GHz (corresponding to a wavelength of about 4 mm), will be able to detect movements of a few tenths of a millimeter. This provides unprecedented accuracy in the automotive field. It allows precise measurement of the distance between the vehicle and the vehicles/objects around it. Another improvement is to make the design more compact, eliminating the need for large antennas.

Self-driving vehicles are one of the most anticipated innovations, and the time until they are fully realized is getting shorter and shorter. As this technology becomes more common, the overall impact (both positive and negative) is sure to be significant. One thing is for sure, we are in for an interesting ride.