Why do we say that now is the era of smart radar?

　　According to MEMS Consulting, the next generation of high-resolution smart radar sensors is essential to achieve L4 and L5 autonomous driving. Of course, cameras and LiDAR play an important role in automotive sensor systems. Radar is particularly useful in long-range detection, bad weather or other sensor failures! Let's follow the automotive electronics editor to learn about the relevant content.

Why do we say that now is the era of smart radar?

　　Innovation in smart radar is just beginning. The current field of autonomous driving requires the creation of a new imaging radar system that can reconstruct the surrounding environment like a lidar, interpret the world around it like a human, and outperform lidar and cameras in terms of all-weather operation and detection distance. The use of engineered metamaterial structures that can form and control beams can create a new radar architecture, and driven by an artificial intelligence (AI) engine, it can achieve the discovery, identification, tracking and classification of objects.

The new generation of radars mentioned above are close to delivery. This smart radar　　that goes beyond digital transmission waves and uses metamaterials and artificial intelligence is a key element in achieving autonomous driving. We discussed with Dr. Bernard Casse, co-founder and CTO of Metawave, some of the obstacles to the innovative solutions of this new radar platform, the market still needs to be cultivated, and the prospects for using smart radars in the next generation of autonomous driving. The following is the relevant discussion content:

　　Q: Why do we say that this is the radar age?

　　Bernard Casse (BC): Stakeholders realize that the timeline for the development of autonomous driving depends on the maturity of decision-making algorithms and the performance of sensors. So far, great efforts have been made to improve lidar and cameras, especially in autonomous driving, but we have encountered performance thresholds in terms of detection range and operating speed. However, other sensors, such as radar, have been relatively less invested. Because radar has always been at a disadvantage compared to cameras or lidar, it lacks the resolution to parse the world. However, automakers now realize that radar is the only sensor that can work at long distances (more than 200 meters) and in all-weather conditions. It just lacks a certain "vision and intelligence".

　　With advanced technology, we can restore vision to radar and embed intelligence. This is the era of smart radar. There are about a dozen startups rethinking various aspects of automotive radar. We expect more startups and companies to tackle this problem.

 Why do we say that now is the era of smart radar?

　　Q: As the CEO of a top automotive company, what do you think users need to know about radar?

　　BC: Elon Musk is right to put a lot of emphasis on radar. It is still the most reliable sensor in the car. Many people, including some automakers, are familiar with traditional old-fashioned radar. When they think of radar, they think of it as a sensor with poor "vision" that can only vaguely "see" a piece of metal (for example, a large piece of metal here and a small piece of metal there). This is true for traditional radar, but the next generation of smart radar is completely different.

　　Point cloud imaging (similar to LiDAR) can achieve image depiction of objects by scanning the beam raster and following certain algorithms, such as identifying road signs, vehicle types, people, lamp posts and more objects. With the increase of visual enhancement and speed, we can embed intelligence in radar. In other words, let the current radar have an enhanced "digital eye" that can learn to recognize the specific features of objects and associate them with the corresponding categories.

　　Q: Could you please tell us more about the importance of artificial intelligence?

　　BC: Artificial intelligence (AI) is key because it is the driving force behind our decentralized intelligence. We are supporters of decentralized intelligence. We believe that radar sensors and any other sensors should have their own brains. The decision-making algorithms in the vehicle should rely on sensor fusion (central intelligence) and individual sensors (decentralized intelligence).

　　This adds another layer of safety to the car. If your car “sees” a bridge using its cameras and/or lidar, now your radar can also say, “Of course, I can see a bridge, too.” The concept of distributed intelligence is even more important for radar, since it is currently the only sensor that can see 300 meters ahead in order to be the first system to provide early warning. Autonomous driving is safer when the autonomous vehicle has multiple sets of data from different sensors, allowing the car to receive information and perceive its surroundings.

　　Q: What is the current development status of radar, especially in the field of autonomous driving?

　　BC: Today’s most advanced radars are 100% digital beam forming (DBF). Although analog radars still excel in terms of performance, the automotive industry has moved away from analog radars due to their high cost, power-hungry architectures, and complexity (e.g., phased array antennas). The automotive industry has always been very cost sensitive, and as the cost of signal processing will eventually continue to decrease, just like other computations on microchips (Moore’s Law), DBF remains an attractive option.

　　But DBF's resolution and speed are still limited. The military is more focused on performance, and still uses analog radar for ballistic missile detection and tracking. For autonomous driving, cost is not and should not be the top priority, performance and safety are the most important!

　　Q: What are the limitations of digital radar?

　　BC: DBF has three main disadvantages, which are somewhat interrelated:

　　1. It is slow. DBF requires milliseconds of waiting time to scan the scene. Signal processing in the digital domain is very important. To achieve an acceptable signal-to-noise ratio (SINR), milliseconds of integration time are required. Processing a series of complex analog and digital circuits and the allocation of digital weights to run DBF will result in higher computing workloads, resulting in slow operation.

　　2. DBF lacks resolution - it cannot see narrow objects or pedestrians. It is not a "true" beamforming architecture, and full beamforming is impractical. This is because it requires a large number of antennas to achieve high resolution. First, implementing many antennas is very expensive and requires installing multiple ports on the radar chipset (unconventional). In addition, this will require a higher number of snapshots to obtain an acceptable SINR (because the omnidirectional noise will overload the system). With the traditional three transmit ports and four receive ports, the resolution is not sharp enough to see pedestrians. It is suitable for cars (with a good trade-off between resolution, SINR and range), but not suitable for non-metallic objects.

　　3. Ghosting: DBF is sensitive to highly correlated signals, that is, DBF will enhance the noise generated by signals from multiple paths and produce ghosting.

　　Q: What innovations are happening in radar right now, especially in the automotive space?

　　BC: First, it is the usual incremental engineering: Tier 1 and tier 2 suppliers focus on pushing the limits of MMIC (monolithic microwave integrated circuit) technology. The real breakthrough innovations are coming from startups: we have seen novel technologies such as Luneburg lens antennas, new algorithms to suppress DBF interference, 3D printed arrays, etc. As we enter the era of smart radar , we expect more radar companies to emerge in 2018 and create more innovations. However, we expect some to be short-lived, while some technologies will survive.

　　Q: What are the obstacles to implementing a new type of radar?

　　BC: The frequency window from 76 GHz to 81 GHz is the first hurdle, as it is all in the millimeter wave range. Components and materials are not readily available, and the components that are available have not been characterized. There are still many controversial areas that need to be resolved. However, we are good at solving this and other challenges, and we have made significant progress in a very short period of time.

　　The barrier of operating at high frequencies is not a game-changer for us, and, in fact, it is actually a boon; it makes it difficult for competitors to catch up, allowing us to maintain our lead. There are still some barriers that we cannot talk about at this early stage because we cannot reveal too much about this platform, but all of them are solvable. So far, nothing is insurmountable. We are just at the beginning of this exciting R&D journey, and although we only started sailing at the beginning of this year, the radar products we are demonstrating and will soon be shipping are very exciting to foresee: L4 and L5 autonomous driving on the road in just a few years.

　　Q: When do you think we’ll see these smart radars installed in self-driving cars?

　　BC: All top automakers predict that we will achieve the highest level of autonomous driving, L5, as early as 2025. We are a fast-paced startup in this exciting industry. We will launch a radar product next year, and we are in talks with all top automakers and strategic partners in the field of autonomous driving. I firmly believe that in order for L5 autonomous driving cars to achieve the required safety level, they must be equipped with the next generation of smart radar , including hardware and software.

　　Q: What was Metawave’s approach to building this new radar platform?

　　BC: Medway International’s unique approach is the use of adaptive metamaterials and artificial intelligence (AI). The overall goal is to create advanced radars with vision, speed, and intelligence. In our approach, we use a fully electronically tunable metamaterial platform to replace expensive and power-hungry phase shifters.

Metawave metamaterial antenna technology

　　From a hardware perspective, this approach is to achieve the same performance as current military analog phase shifters, but without the high cost, complexity, and weight of the components. From a software perspective, we are developing powerful hardware control, signal processing, and decision-making algorithms that are optimized specifically for the hardware.

　　Medway International's new generation radar product is called WARLORD (abbreviated from W-Band Advanced Radar for Long-Range Object Recognition and Detection), and its frequency band is 76~81GHz. WARLORD uses a metamaterial antenna, which can emit a highly directional and controllable electromagnetic beam to accurately detect the position and speed of the target object, and is not afraid of bad weather and environment.

WARLORD radar emits a steerable, highly directional electromagnetic beam

　　This millimeter-wave radar is equipped with a self-developed artificial intelligence engine developed by Medway International: WARLORD Radar AI Thinking like Humans (WRATH), which is embedded with various object recognition and classification algorithms. It can distinguish the attributes of road participants (pedestrians, bicycles, motorcycles or buses, etc.) and provide human-like perception capabilities, which is called "Digital Eye". Because of this feature, WARLORD can also collect 3D radar images.

How WARLORD radar works

    The above is an introduction to "Why is it the era of smart radar in automotive electronics?" If you want to know more relevant information, please pay more attention to eeworld. eeworld Electronic Engineering will provide you with more complete, detailed and updated information.