This sensing technology functions as the “eyes” of an autonomous vehicle. Although a simulator is indispensable for the development of millimeter-wave radar (MMWR) characterized by stable vision capability not influenced by weather, the reality is that no MMWR-specialized simulator exists in the market. OTSL’s AMMWR Simulator is a simulator for MMWR development that can simulate radar reflection more precisely in real time by using radar dedicated mapping and reflection models.
No simulator for MMWR at this time?
The development of autonomous vehicles is accelerating at a rapid pace. Not only car manufacturers and tier one suppliers, but high-tech companies with no previous experience in car development are entering the increasingly robust market.
One of the most important technologies required for autonomous driving is sensing technology, including cameras, MMWR, LIDAR and ultrasound, because it acts as the driver’s eyes, accurately perceiving the environment surrounding the vehicle and providing the information required for control. Advancements in this technology will undoubtedly contribute to making it more practical for use in autonomous vehicles.
Millimeter-wave radar has been under development for the past twenty-odd years in Germany and other European countries. In order to design actual MMWR, however, prior to installing it for the vehicle, a simulator that can reproduce the behavior of radar in a virtual environment is indispensable.
As a matter of fact, there is no such simulator designed especially for MMWR at this time. Simulators manufactures currently use to develop MMWR are more like an autonomous driving simulator than a millimeter-wave radar simulator. This type of simulator basically simulates how an autonomous vehicles would behave when sensors are installed, not for simulating how millimeter-wave would react. So the frequency of MMWR and reflection coefficients are often fixed and cannot be changed in some of the existing simulators.
In addition, this type of existing simulator has two major issues. One is Radar Cross Section (RCS) technology, which is used to obtain reflectivity of target objects such as buildings and trees. But RCS interprets vehicles and buildings as points, and measures reflectivity when the radar collides with the point without size. On the other hand, MMWR reflection differs greatly depending on the material and shape of the object. As long as RCS is used for the simulation, accurately reproducing MMWR reflection based on the shape of detected objects will be extremely difficult.
The other is that camera images are used for the map. MMWR does its sensing at a wavelength that differs from light visible to human eyes. For that reason, a 3DCG map suitable for millimeter wavelength is necessary for accurate simulation.
Because there are no simulators designed for MMWR, developers have managed to design the radars with existing simulators even though those simulators have problems above. But without the ability to accurately simulate the reflection and behavior of MMWR, improving the performance of MMWR remains a hurdle. That is why an innovative simulator is needed to develop a better MMWR.
One solution is the Advanced Millimeter Wave Radar Simulator (AMMWR Simulator) developed by embedded systems and real-time systems developer OTSL Inc. This AMMWR Simulator is designed for development of MMWR for autonomous driving. Target users of the new simulator are car manufacturers, Tier-1 suppliers, and MMIC manufacturers for MMWR. Frequency can be set at 77GHz, 79GHz and 24 GHz; values often used by automotive radar.
Use reflection models to obtain correct behavior of MMWR
One of the advantages of the AMMWR Simulator is to use reflection models.
Conventional simulators use RCS rather than reflection models. As discussed above, RCS greatly simplifies objects such as vehicles, pedestrians, traffic signs, buildings and trees and represents them as “points”. RCS’s reflection data is a numerical value acquired when the radar collides with those points. So it is as if radio waves are radiated from something without size.
However, millimeter-wave reflects quite differently when it collides with a tire and when it collides with a windscreen, for example. So as long as a simulator uses RCS, it cannot accurately simulate the behavior of MMWR when it collides with those automotive parts.
The AMMWR simulator, on the other hand, calculates reflection intensity and received signal power using ray tracing, so reflection models are created to simulate how the ray is reflected on the surface of an object when millimeter-wave collides with it. OTSL has independently developed this reflection model. Using reflection models rather than RCS, the simulator can simulate how millimeter-wave radar will reflect off a front panel, a tire and a bumper.
The AMMWR Simulator can simulate reflection from areas made with different materials in different shapes, such as a windscreen or tire, with greater precision. Source: OTSLAdditionally, simulators using RCS are monostatic, which means a radar transmitter and a receiver are at the same location. Therefore it cannot simulate scenarios in which the transmitter and the receiver are physically separate. Compared to it, antenna location can be set freely in the AMMWR Simulator.
Additionally, simulators using RCS are monostatic, which means a radar transmitter and a receiver are at the same location. Therefore it cannot simulate scenarios in which the transmitter and the receiver are physically separate. Compared to it, antenna location can be set freely in the AMMWR Simulator.
Use 3DCG map dedicated to MMWR
Another important characteristic of the AMMWR Simulator is that it uses 3DCG maps designed for radar rather than maps made with camera images. The 3DCG scenario source code is implemented in the Unreal Engine 4 game engine developed by Epic Games.
In the AMMWR Simulator’s 3DCG map, the reflectivity is already implemented about all the objects in the map such as roads, trees, street lamps, stoplights, and vehicles. When the radar collides with a vehicle, street lamp or stoplight, an accurate reflection based on its reflectivity comes back to the MMWR radiation model. Since the reflectivity changes based on frequency, this 3DCG map includes the reflectivity for each frequency.
A view from the AMMWR Simulator display. The red vehicle is equipped with radar, and the yellow section is the visual field of the radar. The perspective of the 3D image can also be changed freely. The image on the right shows the red vehicle�s driver�s view of the area that is viewed from above in the image on the left. Source: OTSL
MMWR signal processing can also be simulated. Source: OTSLOTSL has also improved 3DCG map precision. The original 3DCG map was created based on the 1m-mesh resolution. Normally, that would provide high enough accuracy, but it’s not enough for MMWR. For example, with 77GHz, wavelength is only about 4mm. In short, precision of millimeter order is required. This special 3DCG map used in the AMMWR Simulator features precision down to less than 2mm.
The AMMWR Simulator features “true real-time” capability
The AMMWR Simulator also uses ray tracing, enabling real-time simulation.
Of course, conventional simulators also simulate in “real time.” But to be more precise, it just “calculates in real time”, not “simulates in real time” because those existing simulators use fixed data for reflection characteristics for simulation. On the other hand, the AMMWR Simulator uses the special 3DCG map and reflection models, so reflection properties change in real time depending on an object or an angle the radar collides with. The AMMWR Simulator has “real-time capability” in the true sense.
Capable of simulating MMWR with different specs
The AMMWR Simulator is capable of MMWR signal processing. First, the radar radiated from a MMWR model collides with an object. Then the simulator calculates power of the received signal that is reflected from the object and returned to the antenna. In the AMMWR simulator, the received signals are calculated by AMMWR signal processing block which based on actual circuit (highly abstracted).
This feature allows AMMWR customers to customize the post-stage signal processing circuit including antenna and MMIC. In other words, you can change characteristics such as those of radiation power, area, and frequency of a MMWR. Therefore, the AMMWR Simulator enables you to simulate MMWR products from various manufacturers and prototyped MMWR products.
There is another advantage. Normally, each MMWR manufacturer implements its own direction detection algorithm in the post signal processing. This algorithm is verified with software at a design stage, and is tested after being implemented into hardware. On the other hand, the AMMWR Simulator can simulate signal processing blocks of a MMWR, so it can verify algorithms by implementing them in virtual space.
MMWR signal processing can also be simulated. Source: OTSLOTSL’s AMMWR Simulator enables tier-1 suppliers to decide radar specification prior to installation and show their own radar advantage on PC.
The simulator also enables Semiconductors companies that design MMIC for MMWR to demonstrate the characteristics of their device under development in conditions close to the actual environment.
The advantage for car manufacturers is that they will now be able to test radar placement and layout by themselves prior to actual installation. “Until now, car manufacturers usually do it with suppliers. So they had to adjust schedule each other and test together. Using the AMMWR Simulator, they no longer have to do that, so they can shorten development time”, said OTSL.
The AMMWR Simulator supports Windows 7, Windows 8 and Windows 10 Professional 64 bit OS. However, as the maximum memory limit for Windows 7 is 196 GB, Windows 10 (maximum memory of 2 TB) is recommended for large-scale simulations.
OTSL offers the radar and sensor simulation framework, “COSMOsim”, for autonomous driving. COSMOsim consists of The AMMWR Simulator and four other simulators for LIDAR, camera, infrared, and ultrasonic. The company provides the AMMWR Simulator and the “ALR Simulator” for LIDAR at this time, and the other three simulators are currently under development.
OTSL�s Simulator Framework COSMOsim (Source: OTSL)Despite the growing importance of MMWR for autonomous driving, until now, there has been no environment for accurate simulation of MMWR because there is no simulator dedicated to the radars. The AMMWR Simulator is a powerful tool that can accurately simulate the transmission and reception behavior of millimeter waves, taking development of autonomous driving sensing technology to a new level.
—Founded in 2003, OTSL Inc. is a technology company that provides high technological capabilities by planning and developing embedded systems, real-time systems, and short-distance wireless communication systems. Its corporate philosophy is "to contribute to the development of human beings by logical thinking and provide highly reliable systems to help customers create products that make people's lives richer." Its business activities include system development, software development, consultation on quality assurance, and provision of educational services. Its aim is to provide a wide range of support to help advanced companies take advantage of the technological benefits of information technology and advance their products and services using their technological advantages. OTSL announced its MMWR solution in October 2017.