|
In many advanced vehicle designs, several radar units are located around the vehicle to provide a complete field of view and allow low- to high-range coverage up to several hundred meters.
At the same time, the semiconductor industry is rapidly moving toward multistatic radars, whose antenna arrays consist of dozens of transmit and receive antennas. Some manufacturers are migrating to full CMOS designs or mixed-signal SiGe architectures to integrate the digital chain into the radar chip.
As a result, radar solutions for ADAS functions and later for autonomous driving have become a cost-effective and irreplaceable solution.
Additionally, machine learning techniques are often used to facilitate the sensor fusion decision-making process to maneuver vehicles in real time on the street. Several global leaders in the digital processing industry are working to implement efficient processors to accommodate the needs of machine learning, such as deep learning algorithms. Some processors are based on GPU architectures, parallel CPUs, or even on dedicated controller units with direct sensor interfaces.
Radar sensors have unique capabilities in measuring the range, radial velocity, azimuth, and size of targets by estimating the time delay, Doppler shift, angle of arrival, and amplitude of the echo signal in the observation area . Some modern radar sensors can also estimate the elevation angle, and the next generation of radars should provide true measurements of the elevation angle.
Determining these parameters simultaneously in complex multi-target environments such as intersection scenarios poses a technical challenge to radar design. To achieve this, radars need to provide high-resolution data, a fact that has encouraged many contributors to report on imaging radars or seek synthetic aperture methods to enhance radar data. All of these requirements place stringent demands on the verification and validation of each radar unit or sensor system to ensure the expected performance.
As radar becomes more complex and intelligent, it is not enough to use direct evaluation of radar signal quality to judge its performance on the street. In addition to conventional tests of its signal phase noise, Doppler resolution, phase reproducibility, temperature stability, output power, receiver noise figure, FM slope and linearity, testing of the overall function is becoming more and more necessary.
Effects caused by integrating the radar inside the vehicle, such as internal reflections from the housing and radome (emblem or bumper), add to this complexity and degrade performance. Therefore, functional testing is becoming a mandatory step for approval by many premium car manufacturers.
The simplest functional test relies on a corner reflector mounted at a specific reference distance in front of the radar. In order to obtain a stable and repeatable test environment, a large anechoic chamber (such as the R&S ATS1000) is usually required to suppress any unknown environmental conditions. Although this sounds simple, this setup can actually only test the detection threshold of a stationary ideal target at a given signal-to-noise ratio level.
It is not possible to test Doppler resolution and the dynamic behavior of the target, for example to verify the tracking and classification process. Therefore, a more realistic setup is necessary to simulate real-life situations. Radars from other moving vehicles must also be included to simulate foreign signals to ensure interference mitigation.
Simulator Implementation Method
The typical implementation method is to receive the radar RF signal and down-convert it to an intermediate frequency, where time delay (range), radial velocity (Doppler shift) and attenuation (RCS) are introduced.
The modulated signal is then coherently up-converted to radio frequency and retransmitted to the radar under test. The radar under test receives and processes the modulation of the signal it originally sent and reports the detected range, Doppler shift and RCS.
Both analog and digital radar echo generators follow the same concept, but their operation on the radar echo signal may differ. While analog echo simulators use delay lines, such as waveguide, coaxial or optical fiber, to delay the signal to a fixed distance, digital solutions also have greater flexibility in dynamically changing the range through programmable time delays.
However, a critical parameter in digital solutions is the delay caused by the associated signal processing. Converting the radar waveform from the analog domain to the digital domain requires at least several digital clock cycles. Since radar signals propagate at the speed of light, each nanosecond of delay will correspond to approximately 15 cm of distance, which cannot be compensated. While analog radar echo generators are used for validation tests and production lines, digital generators are more used in R&D, with the potential to test more complex radar scenarios.
A single radar echo generator can be used to verify tracking algorithms for simple radial motions of targets. This is the case, for example, in many automatic cruise control (ACC) scenarios. To test features such as lane change assistance, the target azimuth must be changed, so the angle of arrival needs to be simulated by the simulator front end.
Due to the huge demands brought about by highly automated driving, the development cycle of automotive radar is shortening. Radar performance, functions and applications are constantly improving. As the number of applications increases, the scenarios in which applications and radar sensors must be ultimately tested also increase accordingly.
Before a function is confirmed, a million test kilometers need to be driven. Given that new sensors and new cars are introduced every year, it is impossible to keep up with driving tests. In addition, decision networks trained with “old sensor” data may no longer be valid, because the training data and classification algorithms rely on the sensors themselves.
This means that a new sensor requires a new training and test dataset, which means another million kilometers of testing. Since future production cars will be highly automated and fully autonomous, we need to find ways to reduce the number of kilometers required for driving tests. For older cars, vehicle-in-the-loop (VeHIL) test benches are available. But for newer cars that rely on information from radar sensors, these test benches must be updated with additional test equipment.
In many cases, the car on the test bench will not even accelerate before the radar is controlled. A radar echo generator and simulating the radar sensor echo through the electronic control unit (ECU) interface is a good starting point. Although software simulation of radar sensors can be comprehensive and meet many needs, it does not truly replicate the real behavior of the radar.
Radar echo simulators test radars, simulating range, Doppler and azimuth. However, current radar echo generators are not able to generate realistic scenarios for the many azimuth and elevation angles that the sensor detects in a normal environment. This is because radar echo generators have a limited number of transmit and receive antennas and therefore cannot simulate changes in the angular orientation of the radar sensor under test. As mentioned before, this is sufficient for simple functional tests or performance tests such as accuracy, detection threshold or resolution, but is definitely not suitable for functional testing of advanced driver assistance systems and autonomous vehicles.
Radar echo simulators may require hundreds of receivers and transmitters to capture, manipulate and retransmit echo signals that are as realistic as typical radar echo signals. In addition to angular limitations, current radar echo simulators cannot simulate distributed targets.
Pedestrians are not just a reflection. They have multiple reflection points, with different velocities for the torso, legs and arms. Vehicles do not just appear as a single scatterer, but rather a distribution of scatterers in both range and azimuth, consisting primarily of a Doppler component. All of these requirements must be considered when generating the realistic radar return signals needed to test tracking, classification and decision-making processes from a scenario and functional perspective.
The image below shows a concept where the radar echo generator is mounted behind a screen with an antenna array. The screen displays a driving scenario, such as a highway scenario, for the camera sensor that supports the driver assistance system.
A fully electronically controlled antenna array with thousands of transmitters and a digital processing backend can be used to simulate complex targets and their maneuvering radar sensors. The sensor is located in front of the measurement system, which receives the radar transmission signal, controls the range, Doppler, radar cross section (RCS) in real time, and transmits the echo signal to a specific antenna in the antenna array, thereby obtaining the azimuth and elevation of the radar under test.
The advantage of this modular approach is that the reflections of the echo signal will be the same as in real life. In this frequency range, large antenna arrays exist and can be used for radar testing, but there are currently no commercial radar echo generation solutions that can generate complex point cloud targets from such antenna arrays.
Testing autonomous vehicles will become more complex as the number of radar sensors increases, as well as as the number of different operating modes and sensor functionality increases. To address these challenges, radar echo generators with a single transmit antenna and a single receive antenna are a good approach, but do not fully meet the needs of future radar sensor and scenario testing.
Antenna arrays combined with digital radar echo generators will be able to more realistically meet the needs of testing radar sensors. As the development of autonomous vehicles, test scenarios, radar sensors and their fusion with other sensors such as laser scanners and cameras continues, OEMs, Tier 1 manufacturers and test and measurement manufacturers must work together to provide solutions for the growing needs.
|
提升卡
变色卡
千斤顶
京公网安备 11010802033920号
