Tencent technology news on September 26, at present, almost all companies in the automotive and technology industry are trying to improve and adjust the century old radar technology to help vehicles drive more safely with less manual participation. Radar is becoming one of the most popular key technologies in the next generation vehicle safety system and automatic driving.
The birth of radar began by accident in the early 20th century. It helped Britain beat back the German air force in the Second World War. It has always supported weather forecasts and allowed air traffic controllers to keep us safe. Today, radar is no longer used only for aircraft and military facilities. In terms of hardware and software, many companies are using radar as an important part of the safety system to detect riders and pedestrians.
The need for this technology is urgent: the number of pedestrian deaths has soared in the past few years, although Americans have reduced their mileage on the road. In addition, the automatic driving level of the new car is getting higher and higher, which can realize the functions of collision warning, automatic braking and blind spot detection, not to mention the automatic driving in the future, which will completely depend on the advanced sensing system. Such a system is also crucial for automobile manufacturers to fulfill their commitment to integrate automatic braking systems into all vehicles by 2022.
In order to reduce and prevent fatal accidents on the road, entities such as Mobileye, a subsidiary of Intel, are developing chips full of micro radar antennas. General Motors recently invested in oculii, a start-up company that makes full use of mathematics and software, using machine learning to shape the signal types used in automotive radar systems. Software company MathWorks is also developing algorithms that allow carmakers to integrate radar and other sensor data into trusted images of the vehicle’s surroundings.
Erez Dagan, executive vice president of products and strategy of Mobileye, said that it is a rapidly changing era for engineers engaged in vehicle sensors. The resolution of the camera used in the car continues to improve, and it can perceive a wider range of natural light than in the past. Radar reflects the laser onto the surrounding objects and “sees” the world in 3D. Now the price is not as expensive as before. Radar is common in the prototype of autonomous taxi, such as waymo, a sister company of Google, cruise of General Motors and Amazon zoox.
When the term radar was born, it was the acronym of “radio detection and ranging”. Since the 1990s, this technology has been used in the first generation safety systems of many vehicles. Automotive radar system has many advantages. When installed in a car, the radar is strong enough to survive years of collisions and temperature fluctuations. They are much cheaper than lidar, are good at measuring the speed of objects in real time, and can resist various adverse weather effects, such as fog and rain, which may hinder the camera and lidar system.
But until recently, radars still have a major disadvantage: their resolution is much worse than other systems, which means that the images they produce are also much blurred in nature.
Oculii’s technology is realized by changing the shape of the radar signal sent by the car radar, that is, the well-known waveform. The physical principle is complex, but by changing the nature of the radar signal (depending on which object it bounces from), it can distinguish objects whose shape cannot be “seen” under other circumstances. Steven Hong, chief executive of oculii, said that as a result, the existing vehicle radar sensors (priced at about $50 each) can generate 3D images of the surrounding environment of the vehicle with much higher resolution. The company’s software will debut in the upcoming lotus lambda SUV, which will be released in 2023 based on radar safety system.
Using the chip manufacturing capabilities of Intel, the parent company, Mobileye is studying a single microchip covering nearly 100 micro antennas. Mobileye says that by using artificial intelligence (AI) software to process the noise signals they receive, its system can do many things, such as identifying pedestrians, at least in the laboratory. This is a function that can only be realized through camera and lidar in the past.
The configuration of camera, lidar and radar will become the standard way to realize various safety systems or automatic driving. There are great differences among automotive technicians, but almost everyone believes that the best solution will be some combination of them. Matthew weed, an engineer and senior director of product management for automotive radar systems, said that even the best automotive radar can only reach the level of the worst lidar system. The system of luminar, an American lidar manufacturer, is superior to radar in most application fields, but the price is as high as $1000.
Wade said that luminar’s radar based systems can justify their high price because they have very good performance and can reduce the insurance cost of drivers by preventing accidents and pedestrian deaths. He added that even if such a system is installed in the car, the radar is a good backup when the radar fails or cannot cope with bad weather.
Mobileye uses radar, cameras and lidar in its state-of-the-art systems. Amnon shashua, CEO of the company, said that although the price of lidar systems has decreased, their cost is still 10 times that of radar due to the complexity of the hardware involved, and may remain at this level in the foreseeable future.
Elon Musk’s Tesla is also going all out, betting that the company can realize real automatic driving on its vehicles by using only cameras. Cameras have the advantage of high resolution, and thanks to the progress of smartphone cameras over the years, they are more affordable and compact. But Mobileye executive Dagan added that for a system that can meet the highest safety standards and even finally achieve full automatic driving, the camera needs backup sensors to prevent it from malfunctioning under different conditions.
Taking fog as an example, this may cause obstacles to the camera and lidar system, which may cause the vehicle to stop when it should not stop. Dinesh Bharadia, an assistant professor of engineering at the University of California, San Diego, said that in a study published in 2020, radar based vehicle sensors had no problem penetrating fog and correctly identifying vehicles hidden in fog.
Dr. baladiya said his team found that one of the keys was the use of multiple radars, which were spaced at least 5 feet (1.5 meters) apart on the vehicle. The number of cameras on the back of our smartphones is increasing, which is also based on the same principle. It is possible to use multiple low-cost radar sensors to create “images” of the environment around the car, just as our mobile phones can use multiple small and cheap cameras and then recombine the images they collect into clearer images.
Rick Gentile once worked on defense radar systems and is now the product manager of MathWorks software. He said that to put all the sensors in a car together to form a single and coherent view of the reality outside the car, we need to integrate all these data. MathWorks is mainly a software company that develops tools to help process data. For example, although the radar may be able to detect a sign ahead, it cannot see the color, which is very important to quickly identify which sign it is.
For the so-called autonomous taxi, the way to fill the capacity gap of each sensor is to use them all. Duggan said that their goal is to achieve “complete redundancy”, so that even if a single sensor fails, other sensors can correctly perceive the world. He believes that this is the fastest way for vehicles to get at least as good senses as people. However, whether these vehicles have enough judgment to really drive safely is another matter.
Before we get a true autopilot, the automakers will have to choose between radar, lidar and camera, or some combination of the three, so as to create a security system that can meet the promise of the automatic braking system as a standard by 2022, and continue to improve these systems.
All three sensors continue to get better, but the cost difference between them leads carmakers to favor one technology or another, depending on the extent to which they think they can make up for their shortcomings with software and artificial intelligence. This has led to healthy competition among manufacturers of security systems, sensors and support software. While all these companies are scrambling to occupy a place in the car, the goal of all these technical experts is to profit by significantly reducing all kinds of traffic deaths while driving. They agreed that this is a goal easier to achieve than fully automatic driving. (reviewed by Tencent technology / Jinlu)