Interview with Fahimeh Rafieinia, UniqueSec AB
The progress of autonomous driving and the deployment of ADAS are creating various tasks related to radar, from radio interference, performance issues and development of performance evaluation methods for state-of-the-art radars such as imaging radars. AEM interviewed with Fahimeh Rafieinia, Chairman of IEEE SA P3116 and CTO of UniqueSec.
Written by Sangmin Han_han@autoelectronics.co.kr
Radar Standardization Needs for AD
Q. What are the current discussions in major countries such as Europe and North America, and how do governments and standards agencies (ISO, IEC, IEEE, etc.) view them?
A. Fidelity of sensors is one of the critical aspects around sensors (cameras, lidars and radars) worldwide. It is also a matter of discussion by different standardization bodies to evaluate the sensors’ perception versus reality. IEEE P3116 is an example of such bodies for evaluation of automotive radars’ performance. Respectively, IEEE P2020 and P2936 address vision sensors and lidars.
With the rise in the number of radars on the vehicles especially for higher automation levels, congestion of the spectrum and radars exploiting larger bandwidths, another important issue is how much we can trust these sensors when they are operating in close vicinity, causing interference for one another. In IEEE P3116, we discuss different sources and types of interference and our goal is to provide methods to evaluate the performance of automotive radars under such interference.
ISO has also two ad-hoc groups concerning radars. They are mainly focused on cooperative interference mitigation techniques and radar
performance. There are activities within ISO with regards to safety functions as well.
Artificial Intelligence (AI) and AI-based methods in the automated driving context are facing questions about trustworthiness and explainability. IEEE SA has several working groups on different aspects of AI, including ethical AI, which is quite important when the lives of humans depend on decisions made by AI.
On the national level, different countries such as Sweden and Germany have created AI development and superiority plans. These plans extend to the autonomous vehicles domain as well. However, the industry is well ahead of standardization entities and governments in deployment of AI. While AI is now being utilised in automated driving, sensing and processing, there are no standards considering these use cases yet.
Standardization is not enough by itself either. Adoption of a standard through regulation is a necessary step for a standard to actually be implemented in real life.
Transition between automation levels is also a debatable topic in the automated driving context. With the automated levels characterised as being hands-off and/or eyes-off the road, the challenge is how to define the operation design domain (ODD) for level 3 or 4 when the system needs to judge and decide by itself if it is no longer in the defined ODD. In level 5, ODD needs to include a countless number of driving situations, demanding a huge amount of considerations in the system design.
All these issues place limitations on the deployment of automated driving systems and therefore the implementation is being narrowed to robotaxis for now.
Q. What are the possible problems caused by radio interference of Radar for vehicles?
A. Radio interference can cause radars to see false targets that are stronger than real targets, meaning a high false alarm rate. It can also increase the noise level at radar receiver so the target cannot be detected, i.e a higher mis-detection rate. The radar resolution can also deteriorate, especially at the edges of the FoV, as a result of interference.
One issue being raised in IEEE P3116 is that with many radars using the air interface, the use of radars with higher emitted power might be encouraged since they naturally perform better against interference — Although it is not really a good direction for the future and not sustainable either. Another point is the necessity to have a standard approach for characterising the impact of interference from various sources and directions on different radars-under-test.
Q. For outdoor radars for vehicles, the FMCW method is most commonly used in the 77-79 GHz range, how stable is it for radio interference?
A. E-band, where most automotive radars operate, are intrinsically good at handling interference because the wavelengths are short and losses are high. However, in frequencies between 76-77 GHz, used by many legacy radars, the spectrum is getting crowded. Most radars now use bandwidths of below 500MHz without any interference avoidance mechanism. But radars are moving towards more complexed FMCW modulations, such as stepped FMCW, or use mechanisms such as hopping frequency to avoid or reduce interference effects.
She would like to invite Korean stakeholders in the area of automotive radars and autonomous driving to join IEEE the P3116 Working Group to introduce the Korean perspective on the safety and performance of radars. She said they can reach out to me directly for coordinating their membership in the working group.
Q. Is it necessary to develop new technologies to avoid or reduce radio interference? What technology is seen as promising? What is radar's potential? You said radar is the most reliable sensor for level 4 and above, why?
A. Yes, absolutely. There are several interference avoidance technologies, including:
- Use of RadCom (Radar and communication) to avoid interfering with neighbouring radars,
- Use of compass, so that radars in opposing moving directions use different frequency bands,
- Using mechanisms such as listen-before-transmit, to minimise the risk of interfering radar signals,
- Use of spatial nulling in signal processing to avoid harmful impact of the interference from certain directions.
These are examples of approaches suggested by research. Radar communication is perhaps the most advanced amongst these, however, it takes more work and perhaps also standardization around it before it could be deployed. I believe there will be successful schemes out there before interference becomes a major threat to automotive safety.
In L4, we need sensors which are available in all weather and lighting conditions. As you know, vision and Lidar have restrictions in severe weather conditions and this limits their usage for L4. On the other hand, radars offer high performance and availability regardless of weather conditions.
Q. What regulations, standards, and regulations do you expect for radio interference in vehicle radars in the future? Is there a roadmap for this? The activities of the IEEE SA P3116 working group include radar frequency interference problems. What is the current progress of this standard? You said this standard would be developed in five years. Isn't it too long?
A. I think frequency regulators such as ITU, ETSI and FCC should look into air interface usage and congestion by automotive radars considering the growing number of radars on roads. Effective interference mitigation methodologies should be standardised for the purpose of interoperability. The ISO ad-hoc group, which I mentioned earlier, works on proposing cooperative interference mitigation methods.
IEEE SA P3116 mostly focuses on analysis and evaluation of radar performance under interference. It also looks for proper ways to introduce interference specially from moving entities (namely other automotive radars) to the radars and evaluate radar capabilities to detect the actual targets. So, we have two sub-groups focused on interference and testing.
One of complexities we are working on in IEEE P3116 is the evaluation of MIMO and imaging radars in a way that resembles the real-life application of these radars. We are discussing simple in-lab scenarios, addition of advanced targets and interference emulation to the lab setup as well as simulations covering complex real-life scenarios. As you can imagine, there is no easy solution. I really hope that the requirements for testing radars in this work will drive and shape the progress of radar test methods and radar target simulation technologies towards the safety goal for AD.
The timeline for this standard is five years, and it has started near the end of 2021. So there is a long way.
On the one hand, it is a very extensive task to develop a standard for automotive radars and on the other hand, it requires contribution and acceptance from experts of the industry for such standard to succeed. I think our work can be accelerated if this standard is demanded by national authorities, for example. Their work is to establish the infrastructure and regulations for test and type approval of sensors in level 3 and beyond.
Q. How can we test this radar, including the performance, simulation, and virtualization required by autonomous driving?
A. Testing of automotive radars, specially those which are used for autonomous driving, is a difficult subject. In a more realistic test setup, the radar, radome and even the vehicle bumper are included and the target’s perception is created over-the-air using a radar target simulator (RTS). Such RTS should be able to provide perception of:
- static and moving objects,
- at various distance, speed, angle and radar cross section/size,
- multiple targets with arbitrary trajectories,
- environmental elements such as rain, snow and fog,
- interference from static and moving sources coming from different directions,
- Objects at short and long distances.
There are two major technologies in the market for radar target simulation (RTS). The most common is delay-based which originates from old DRFM technology. And the new trend is the frequency-domain RTS, invented by Uniquesec.
Q. You described Uniquesec solution for imaging radar MIMO radar test in IWPC. Please summarize this.
A. ASGARD2 is our state-of-the-art solution for emulation of target's angle perception. It uses MIMO methods as well as our patented technology for radar target emulation in frequency domain rather than time. The frequency-domain signature generation is also featured in our first generation product, ASGARD1.
ASGARD2 emulates the perception of angle in addition to distance, speed and RCS for multiple arbitrary moving targets without any rotation, mechanical movements or the need for thousands of antenna elements in a huge array. It is the only product that works in near-field of the radar, allowing it to be ultra-compact and specifically suitable for MIMO and imaging radars, where the far-field distance of the radar is quite huge (several meters). All the other RTS solutions for automotive radars rely on operation in the far-field region, which imposes requirements on the size of the testing chamber that are hard to meet.
Vehicle-in-loop testing and end-of-line calibration only allow for RTS systems with compact form-factor due to space limitation.
The biggest advantage of ASGARD in comparison with other products in the market is the possibility of emulating targets at extremely short minimum distances, making it suitable for testing bumper to bumper traffic and safety functions such as AEB and side assist. Then is the huge number of targets that ASGARD can emulate, thanks to the flexibility that working with the spectrum offers. Adding to the number of targets at various angles with ASGARD2 does not need more RTS antennas, so it is possible to emulate complex test scenarios for high resolution MIMO radars. Besides, our solution is capable of generating multiple interferers on moving vehicles in the same scenario with all other moving targets.
ASGARD2 emulates the perception of angle in addition to distance, speed and RCS for multiple arbitrary moving targets without any rotation, mechanical movements or the need for thousands of antenna elements in a huge array. It is the only product that works in near-field of the radar, allowing it to be ultra-compact and specifically suitable for MIMO and imaging radars, where the far-field distance of the radar is quite huge (several meters).
Q. Have you been in touch with Korean officials since your interview article was published in May?
A. We have not directly been in contact with Korean authorities regarding the standard. But a company in Korea called TAENG engineering is trying to market our products in Korea. I believe they are in contact with a national lab for this product.
I would like to invite Korean stakeholders in the area of automotive radars and autonomous driving to join IEEE the P3116 Working Group to introduce the Korean perspective on the safety and performance of radars. All OEMs, tier-1s, tier-2s, independent labs, research institutes, government agencies and standardization bodies are very welcome to contribute. They can reach out to me directly for coordinating their membership in the working group.
AEM_Automotive Electronics Magazine