“Why is the radar scene simulator unanimously recognized by the company and customers? A few days ago, Zhu Xiaoyue, Business Development Manager of the Greater China Region of Keysight’s Automotive and New Energy Division, and Ma Jianrui, Business Manager of the Greater China Region of Keysight’s Automotive and New Energy Division, gave a detailed interpretation.
The launch event was held in 6 locations and will soon be moved to the CES site-the radar scene simulator launched by Keysight Technology recently received such a high standard treatment.
Why has the radar scene simulator been unanimously recognized by the company and customers? A few days ago, Zhu Xiaoyue, Business Development Manager of the Greater China Region of Keysight’s Automotive and New Energy Division, and Ma Jianrui, Business Manager of the Greater China Region of Keysight’s Automotive and New Energy Division, gave a detailed interpretation.
The popularization of automobile autonomous driving technology is inseparable from test technology
At present, the automotive market is evolving towards electrification and electrification. Among them, electrification, represented by ADAS (AD) technology, is inseparable from the development of communication technology between cars and the application of various sensors.
In order to realize the future of autonomous driving, sensor technologies including CMOS image sensors, millimeter-wave radars, and lidars are required, powerful AI processing functions, and enough data for learning, and complete testing and verification. Process, so that enterprises and consumers have enough confidence to adopt these new technologies.
Talking about the testing and verification process, Zhu Xiaoyue said that some research institutions have shown that achieving safe autonomous driving from scratch may require as many as tens of billions of miles of road testing, but this is not realistic.
In order to solve this problem, the industry currently mainly adopts two methods for simulation testing. A quick way is software-based simulation. This method directly combines scenarios, vehicle dynamics and other models to send the simulated sensor ideal data to the ECU for simulation and improvement of algorithm functions.
Zhu Xiaoyue said that the advantage of this method is that it is very efficient and supports rapid iteration, which is very suitable for early import. However, the pure software method also has its shortcomings. After all, as the development enters the middle and late stages, more real physical world scenes need to be introduced, so the current practice of most companies is the actual road test in closed places. It is difficult to construct complex scenarios for actual drive tests, such as rain or snow or sudden emergencies, resulting in the dilemma of low test efficiency and high cost.
Rather, the radar scene simulator of German technology is dedicated to filling the gap between software simulation and road testing. It not only has the flexibility and speed of software simulation, but also combines the authenticity and complexity of road testing.
Mr. Thomas Goetzl, Global Vice President and General Manager of the Automotive and Energy Solutions Division of Keysight Technologies, said: “Keysight’s radar scene simulators provide car manufacturers with a groundbreaking solution through full-scene rendering in the laboratory. The road test can be completed within.”
Features of the new radar simulator
The radar simulator is called a truly “disruptive” product by Janet Ooi, head of Autonomous Vehicle Solutions at Keysight, with the following features:
It can present up to 512 targets with ±70° horizontal FOV and ±15° vertical FOV. It can generate static and dynamic targets from 1.5 meters to 300 meters at a speed of 0 to 400 km/h, and supports short-range, medium-range and long-range Millimeter wave radar.
Use a fixed RF front-end to solve multi-target and multi-angle scenes, provide repeatable angle of arrival (AoA) accuracy, can simulate complex and radio-intensive urban scenes through real interference tests, and use 3D point clouds and multiple reflections Improve the method of detecting and distinguishing objects.
Cheryl Ajluni, Marketing Director of Keysight Automotive Solutions, said: “Automakers know that automated driving algorithms are very complicated to test, and if the test is not in place, it may cause product quality problems. Keysight’s new radar scene simulator is at a leading level. Resolution means a major leap for ADAS/autonomous driving system developers who value safety.”
How to solve the pain points of simulation?
Zhu Xiaoyue introduced the current pain points of radar simulation, compared the difference between Keysight’s new radar simulator and the existing radar simulators on the market, and elaborated on the many advantages of the new generation of radar simulators.
The first is the limitation of the number of simulated targets and the viewing angle, the second is the limitation of the simulation distance, and the third is the lack of target simulation accuracy. These three limitations make traditional hardware radar simulations unable to effectively simulate the real world.
Keysight uses fixed 512 RF front-end modules to support a maximum of 512 simulation targets, and achieve a horizontal ±70° and vertical ±15° FOV, which can meet almost all millimeter wave radar indicators on the market .
In terms of distance, it is difficult to achieve close-range simulation, and the previous products can only support a distance of about 4m recently. However, close-range detection is an indispensable part of the ADAS system, especially for scenarios such as emergency avoidance and pedestrian detection.
Comparing the traditional simulation platform with the Keysight simulation platform, the Keysight platform supports more targets and supports 1.5m ultra close range.
Regarding the accuracy issue, Zhu Xiaoyue cites that the radar needs to distinguish various objects in detail, such as bicycles, cars, and even foreign objects dropped from the vehicle in front, which requires the support of high-precision simulation.
As autonomous driving moves to a higher level, some corner cases begin to appear, and the machine begins to be helpless in the face of some special scenes that have not been learned. This has also become a stumbling block for the mass production of high-level autonomous driving. And the simulation of these corner cases is exactly the advantage of Keysight products.
Black technology supported by 20 patents
Ma Jianrui said that the radar scene simulator only supported 20 targets in July, and then quickly broke through to 512 at the end of this year. This deeply reflects the strength of Keysight in the speed of technological innovation.
As shown in the schematic diagram, there are 512 RF front-ends inside the dark room, each of which is a group of 8 RF front-ends, and each RF front-end component is developed based on Keysight’s own design and Fab.
In fact, Keysight has been relying on its differentiated ASIC bottom design team to complete important breakthroughs in many markets. For example, its UXR series oscilloscopes have used the company’s large amount of indium phosphide technology as support.
Zhu Xiaoyue said that this kind of pure solid-state RF front-end simulation can increase the number of multi-target simulations by tens or even a hundred times. At present, there are also radar target simulation schemes combined with robotic arms on the market. However, due to the delay of the movement of the robotic arms, the action design is complicated and difficult to reproduce, and the real-world effects cannot be perfectly presented. Another solution is to use the antenna array form, but the back end also only has a few RF front-end simulators, and it is also impossible to simulate multiple target scenarios at the same time.
Compared with traditional simulation, the support of 3D point cloud can better simulate changes in the real world
In addition to hardware innovation, Keysight has always attached importance to innovation in the software field, and has taken software as the company’s key development direction.
Zhu Xiaoyue introduced that the radar scene simulator fully reflects the importance of Keysight’s software. On the one hand, the simulator software of the product itself is very powerful, and the simulator is also part of the large platform of Keysight’s automated driving simulation system. The demand for sensor fusion simulation continues to emerge, and the platform-based strategy of synchronous simulation also caters to the needs of customers.
In addition, Zhu Xiaoyue also said that in order to protect users’ investment, Keysight’s radar scene simulator supports a variety of accessory options, which customers can flexibly customize according to their own needs.
Radar scene simulator supports a variety of additional options
Ma Jianrui said that Keysight has previously released a C-V2X-based autonomous driving simulator, combined with a millimeter wave radar simulator, to form a complete autonomous driving simulation platform-Autonomous Drive Emulation ADE. On this platform, Keysight can provide customers with the most complete simulation methodology in the form of solutions.
The platform is jointly built by Keysight, IPG Automotive and Nordsys through many years of cooperation. The ADE platform provides time synchronization input for vehicle sensors and subsystems such as Global Navigation Satellite System (GNSS), Internet of Vehicles (V2X), cameras and radar DD by rendering predefined usage scenarios to train ADAS and AD software. With the open platform of Keysight’s Autonomous Driving Simulation (ADE), automakers and their partners will be able to concentrate on developing and testing ADAS/AD systems and algorithms, including sensor fusion and decision-making algorithms. The platform can also be integrated with commercial 3D models, hardware-in-the-loop (HIL) systems, and existing test and simulation environments.
In 2016, Keysight changed its organizational structure, focusing on customer-centric differentiated solutions, uniting various internal resources within the company, and breaking the original divide and conquer situation. Specifically, the underlying technology is uniformly provided by KEYSIGHT Labs, the Center of Excellence is responsible for supporting hardware and software, while the Communication Solutions Division and the Electronics Industry Solutions Division are detailed to the vertical market, and through in-depth communication with customers, Obtain demand through close communication with the industry. Under this kind of organizational change, it is easier to find customer pain points and respond quickly to provide a complete software + hardware + service solution.
Through the introduction of the special page of Keysight’s official website, we can also see the research and development process of the radar scene simulator, which involves the roles of IC research and development, product development, research and development scientists, software research and development, system integration and other roles, as well as cross-departmental collaboration.
As introduced by Jay Alexander, Senior Vice President and CTO of Keysight Technologies at the 2020 Investor Summit, as of April 2020, Keysight Technologies has developed 88 solutions to meet different directions including interconnection, electric and autonomous driving. s solution.