This paper describes Altair's solution for ADAS through a combined workflow between Altair Feko and Altair WinProp. More specifically, the procedure to set up different ADAS scenarios, analyze and post-process using the FMCW chirp are explained in detail.
Altair EmployeeIntroduction
RADAR (Radio Detection and Ranging) systems find their use across various sectors including military, aircraft and maritime navigation, space exploration, security surveillance, and more. In recent years, there has been a growing interest in its application within the automotive industry to mitigate road fatalities through Advanced Driver Assistance Systems (ADAS), enhancing passenger safety features such as Adaptive Cruise Control, Blind Spot Detection, Lane Change Assist, Rear Crash Warning, and Stop & Go capabilities. Currently, ADAS systems utilize components operating at mmWave frequencies, typically 24GHz or 77GHz. However, due to the constraints of bandwidth and resolution at the lower frequency, 77GHz is largely favored in the automotive industry.
RADAR systems can be categorized as either Continuous Wave (CW) or Pulsed Wave (PW), depending on the nature of the transmitted signal. Due to the advantages such as higher resolution, lower peak-to-average power ratio, and absence of blind spots, CW RADAR is frequently favored over PW RADAR. However, PW RADAR often excels in applications that necessitate longer-range detection capabilities. Considering that longer range provides a narrower field of view compared to broader field of view with shorter range, the selection of a suitable RADAR system should align with the specific application requirements.
The CW RADAR can use either a modulated or an unmodulated signal. An unmodulated CW RADAR can only detect the presence of a target but cannot measure the range or velocity. This limitation arises because there is no reference point to determine the delay between transmitted and received signals. In contrast, a modulated CW RADAR can accurately measure the range, velocity, and angle of a target. Modulation can be performed over amplitude or frequency of the signal. Due to its numerous advantages, we focus on Frequency Modulated Continuous Wave (FMCW) RADAR operating at 77GHz in this article.
FMCW RADAR
The operation of FMCW RADAR can be understood through the block diagram in Fig. 1, which comprises of a transmitting section and a receiving section. The transmitter unit involves Radio Frequency (RF) signal generation, f(t), via function generator and Phased-Locked Loop (PLL), to multiply or divide frequency. Often a voltage-controlled oscillator is used instead of PLL for generating the frequency range by tuning the voltage across a capacitor. The receiver unit involves post-processing of the received signal and will be discussed in detail.
Fig. 1: Complete representation of FMCW RADAR.
When a RADAR receives a signal from a moving vehicle, it encounters differences in both time (Δt) and frequency (fd). Therefore, the echo received is represented as f (t - Δt) +/- fd. This signal has undergone attenuation and contains noise, due to its interaction with the atmosphere. Hence, a robust low noise amplifier is required to amplify the echo for further analysis. Next, a mixer unit is employed to calculate the difference between the transmitted and received signals, producing an Intermediate Frequency (IF) signal. The mixer down-converts the received signal, by determining the difference between the transmitted, f(t), and the received waveforms, f (t - Δt) +/- fd, to generate IF. This step is necessary to promote easier construction of RF components at low frequency. This IF signal then undergoes filtering through a low pass filter or band pass filter to eliminate unwanted noise and harmonics.
Post-processing is the final and most critical stage in achieving accurate results in FMCW RADAR analysis. A key component of this process involves performing Fast Fourier Transform (FFT) to extract information regarding range and velocity. Performing first FFT, known as Range FFT, on the obtained IF from, lets say two targets, leads to two peaks in the frequency domain, each separated by their Doppler frequency. However, the FFT being a complex operation, includes information about both phase and amplitude. Range FFT focusses only on the absolute part of FFT and may give incomplete information about the targets. For instance, if multiple targets are present at the same distance from the transmitter, they won’t be distinguished based on Range FFT. This leads to the need for Doppler FFT, which segregates range and velocity information by considering both phase and amplitude. In certain extreme cases, targets may have the same range and velocity. In such scenarios, a third FFT, known as the Angular FFT, becomes necessary to determine the angle of objects relative to the RADAR's position. This step helps resolve ambiguities and provides comprehensive spatial information about detected targets.
Altair Solution for ADAS
Altair offers a complete solution for ADAS through a combined workflow between Altair Feko and Altair WinProp. The design process of the RADAR transmit and receive arrays using Altair Feko is beyond the scope of this article, but the procedure to set up different ADAS scenarios, analyze and post-process using the FMCW chirp will be explained in detail. The two modes of specifying the RADAR characteristics for FMCW post-processing in WinProp are ‘Set Radar Parameters’ and ‘Design Radar Parameters’, as depicted in Fig. 2.
Fig. 2: FMCW RADAR postprocessing options in WinProp.
Set Radar Parameters: This option allows the user to set the chirp characteristics, including Chirp Duration (T_c (µs)), Sweep Bandwidth (BW) in MHz, Number of Chirps (N) and Frequency Bins (8/16/32/64/128/256/512/1024/2048).
Design Radar Parameters: This option extends the capability to design a RADAR by specifying Max. Velocity (m/s), Velocity Resolution (m/s), Range of Interest (m) and Range Resolution (m).
Once the RADAR parameters are set, one can request for the output in terms of a ‘Doppler-Range Heat Map’ or ‘Angle-Range Heat Map’, for the selected windowing function. The paper attached to this article presents three different scenarios with varying positions of vehicles and the surrounding environment and explain the FMCW post-processing results for each scenario.
Scenario I
The first scenario is a basic time-varying indoor scenario including three vehicles, CAR-A moving at 10 m/s in the x-direction, TRUCK-B moving at 5m/s in the x-direction, and CAR-C moving at 10 m/s in the opposite (-x) direction, as depicted in Fig. 3. Additional details like the road underneath the vehicles and the side railings are also considered in this scenario.
Fig. 3: Position of the vehicles and their respective velocities in Scenario I.
The FMCW RADAR antenna is positioned at the bumper of CAR-A. This antenna position allows the radiation pattern, a full view of targets in front, without much interference of the car itself. The radiation pattern of the mounted antenna determines the extent of interaction of transmitter section with the surroundings. CAR-A is also marked as a ‘Mobile Station’ with the same radiation pattern to work as a receiving unit. At time step of 0.5 seconds, CAR-A is positioned 18.65 m from Truck B and 29.6 m from CAR-C. The relative velocity of Truck B is 5 m/s with respect to CAR-A, while relative velocity of CAR-C is 20 m/s as it is travelling in the opposite direction. Due to the location of CAR-A and radiation pattern of on-board antenna, echoes from side-railings (stationary targets) are also tracked. The distance of CAR-A from the right railing (measured with respect to travelling direction of CAR-A), is 1.5 m, while from left railing is 5.5 m.
The FMCW post-processing yields a Doppler heat map illustrated in Fig. 4, where power normalized with respect to the peak, plotted between relative velocity (m/s) and range (m). Higher normalized power values indicate stronger reflections from targets. The strength of reflections can depend on several factors including distance, material of target, etc., as can be noticed with the blue color of the site at 0.5 seconds.
Fig. 4: FMCW post-processing results for Scenario I at 0.5 seconds.
Scenario II
The second scenario provides a more detailed version of the first scenario, incorporating elements such as vegetation, sidewalks, pedestrians, and road dividers. The interaction of rays in this scenario becomes complex due to these detailed features, making it more representative of real-world environments. Also, the different types of materials for various objects offer distinct transmission and reflection losses, which WinProp can handle efficiently. In this scenario, as depicted in Fig. 5, CAR-A is moving at 10 m/s in the x-direction, where as CAR-B is moving at 10 m/s in the opposite -x-direction. Antenna properties remain the same as in Scenario I and is positioned at the bumper of CAR-A.
Fig. 5: Position of the vehicles and their respective velocities in Scenario II
The Doppler heat-map from the FMCW post-processing helps identify the locations of targets easily. As illustrated in Fig. 6, at time step of 0.5 seconds, the position of vehicles is changed based on their speeds, with CAR-A 27.62 m from CAR-B, at a relative velocity of 20 m/s.
Fig. 6: FMCW post-processing results of Scenario II for the position of the objects at 0.5 seconds.
Scenario III
To enhance realism in the indoor scenario, a hybrid approach combining indoor and urban elements has been implemented in Scenario III. The urban component incorporates an Open Street Map (.osm) file, which outlines the building layout of a specific region. This scenario includes additional details such as cars, streetlights, and traffic lights to enrich the simulation environment, as shown in Fig. 7.
Fig. 7: Time-varying hybrid scenario.
In general, it becomes challenging to identify the targets with the increased complexity of the scenario because of the complex network of ray interactions with the surrounding objects. However, due to the heat plots generated with respect to relative velocity and range in WinProp, it becomes very convenient to map the targets with respect to the surrounding environment, as illustrated in Fig. 8.
Fig. 8: FMCW post-processing results for Scenario III, for the respective position of the objects.
Apart from the various scenarios described above, the attached paper also describes and compares the propagation models in WinProp, namely, the Shooting and Bouncing Rays and the Standard Ray Tracing methods.