Autonomous driving – still on the road to improvement
Latest update time：2021-03-16
Reads：
There is no doubt that autonomous driving is reshaping the role of cars in our lives. At its current stage of development, in addition to the many unresolved issues at the legal and liability levels, technical challenges remain severe.
For the automotive industry, the future of autonomous driving is a given, and the only variable is when it will be deployed on a large scale. At present, the market should be more concerned about its "application scenarios" and "how to implement" issues.
For several years, automakers, Tier 1 suppliers and semiconductor manufacturers have made great strides in self-driving vehicle technology.
However, all mass-produced models have not yet reached Level 3 (autonomous driving under limited conditions).
The first mass-produced model equipped with a highly automated driving system is expected to hit the road in Germany in 2021, but it will only be able to drive on specific types of roads, such as autobahns, and will also need to meet suitable climatic conditions. In order to improve the adaptability of autonomous vehicles in non-ideal environments, sensor groups, computing units and software algorithms need to be continuously optimized.
Automotive Industries (
hereinafter referred to as "AI"
)
recently
conducted an exclusive interview with
Robert Schweiger, Director of Automotive Solutions
at Cadence
.
The following is the content of the interview:
Robert Schweiger
Director, Automotive Solutions, Cadence Corporation
AI: Why is it so difficult to achieve Level 3 autonomous driving?
Robert Schweiger:
In the past few years, we have witnessed technological competition among various companies and countries. It is obvious that each manufacturer hopes that its self-driving mass production model can be approved by its own government first.
Although there are still challenges that have yet to be overcome, technological innovation and advancement are clearly ahead of the formulation of government regulations. The relevant agencies have not yet issued any legal framework regulations to allow car manufacturers to use these on the road under certain conditions. A new technology. For this reason, Audi's Traffic Jam Pilot autonomous driving system has not been installed on its A8 series production models.
In addition, self-driving cars face many other regulatory hurdles such as liability and insurance.
Now, the competition in the field of autonomous vehicles has penetrated into the legislative and regulatory level of governments of various countries.
AI: Speaking of new technologies in the field of autonomous driving, can you share the key trends in hardware platforms?
Schweiger:
Autonomous driving systems are likely to be first used in robotaxi and luxury models, because OEMs of high-priced cars will be able to more easily afford the increased cost of autonomous driving systems. Looking to the future, the most promising application areas should be mid-size cars and economical electric cars. Key requirements for hardware platforms in this market segment include performance per watt, scalability, driving safety, information security, and cost.
Using 10 different electronic control units (ECUs) simultaneously means that it is difficult to meet the stringent performance and cost requirements per watt, so the only way to deal with this is to use highly integrated solutions based on high-performance system-on-chip (SoC).
In 2019, Tesla released its first fully autonomous driving (FSD) computer (HW3.0), as well as supporting self-developed software and machine learning environment. What is even more commendable is that as the first OEM manufacturer to develop its own SoC, Tesla's SoC chip can fully match its own system needs.
The SoC released by Tesla is composed of two AI processor cores, with a maximum performance of 144 TOPS/72 watts (2 TOPS/W), which is still the industry benchmark.
Most other autonomous driving platforms on the market do not even reach 1 TOPS/W! The goal of the next generation SoC should be hundreds of TOPS and a better TOPS/W ratio, and achieving this can only rely on the most advanced process technology on the market.
If an automaker really wants to enter the commercial market for autonomous vehicles, they should seriously consider developing SoCs based on their own needs and build an autonomous driving platform that optimizes performance and cost.
AI: Will the massive data generated by smart cars create new demands on memory?
Schweiger:
Even though hands-free autonomous driving systems (Level 3) are expected to be used in specific road environments in 2021, I think the real value of autonomous driving lies in whether it can also support normal operation on construction roads or in adverse weather conditions. To this end, we need higher-precision sensor modules to improve the robustness of the system.
This is why similar systems generate huge amounts of data. When our autonomous driving evolves from Level 2 to Level 5 fully autonomous driving, it is expected that 3 GB to 10 GB of new data will be generated every second. Even a Level 2 autonomous car generates up to 1 GB of data per second.
Regarding the memory requirements of the new generation of ADAS computing units, we have observed the development trend of the latest memory standards.
As an IP provider, Cadence already fully supports these standards.
This changing trend in the automotive industry can be understood from three aspects:
(Figure: Development trends and challenges of electronic and electrical architecture)
AI: The transmission of massive data will significantly affect in-vehicle network systems. What progress has been made in this field?
Schweiger:
This topic is also very important. Regarding sensor raw data fusion, the two current directions of debate are whether distributed processing or centralized processing architecture is needed. In order to connect high-precision sensors such as cameras or radars to computing units, we need high-speed protocols such as Automotive Ethernet or MIPI CSI-2.
Each current network standard has its limitations. OEMs have already put 1 Gbps automotive Ethernet PHYs into production, supporting data transmission over 15-meter cables. However, insufficient bandwidth is still the biggest obstacle, and even full HD video at 60 frames per second (FPS) cannot be supported.
MIPI CSI-2 combined with the D-PHY v2.1 interface can support up to 6 Gbps per line on short channels (maximum data rate guaranteed within 15 cm). However, due to the limited operating range of the PHY, the maximum data rate for long-distance channel models (1-4 meters) is less than 4.5 Gbps. Long-distance, high-speed interconnections between systems require the use of low-voltage differential signaling (LVDS) bridging.
The MIPI Alliance recently released the MIPI A-PHY specification, which aims to support a data transmission rate of 16 Gbps on cables within 15 meters. It is mainly used to connect high-precision sensors (endpoints) and central processing units.
Looking to the future, the OPEN Alliance SIG and IEEE have initiated the development of multi-gigabit Ethernet standards to support data rates up to 10 Gbps, which is the perfect choice for communication between the Ethernet backbone and ECUs. Of course, 25G PHYs are also on the agenda, but the specific implementation will take longer.
Needless to say, the availability of these two PHYs will have a huge impact on automotive network architecture. I believe that the MIPI standard, which is primarily optimized for one-way data transfer, will be used for endpoints such as sensors and displays; while automotive Ethernet is ideal for high-speed communication between different domains.
AI: How does Cadence help customers develop highly complex systems?
Schweiger:
Cadence has a deep foundation in automotive intelligent system design and is a trusted partner. Cadence can provide all the EDA tools, design flows, silicon-proven interface and processor IP, high-performance cloud infrastructure and leading design services required for automotive intelligent design to help customers create complex automotive SoC chips and smart sensors, and High performance sensor fusion unit.
Custom SoC design is still a new field for traditional OEMs and first-tier suppliers. Cadence's design services team can help customers implement complex SoC designs and strengthen the skills required by the team. With our assistance, customers can gradually master the adoption of advanced chip design processes and eventually execute complete designs independently.
Finally, at the system analysis and design level, Cadence also provides a complete set of computing software for chips, packages, RF modules, motherboards, as well as networks and systems. And our latest system analysis tools such as electromagnetic interface/electromagnetic compatibility (EMI/EMC) and full 3D thermal analysis can simulate complete systems.
For example, customers can apply the Cadence Celsius Thermal Solver integrated with FEA-CFD to analyze the heat flow of the driving platform while taking into account the influence of electronic components and mechanical chassis. In addition, customers can use the Cadence Clarity tool suite to simulate ECU EMI/EMC in multiple ways: either individually, or on a complete automotive platform using the Cadence Clarity 3D Solver for near-field simulation, or using Cadence Clarity 3D Transient Solver performs far-field simulations.
About Cadence
With more than 30 years of expertise in computing software, Cadence is a key leader in the electronic design industry. Based on the company's intelligent system design strategy, Cadence is committed to providing software, hardware and IP products to help electronic design concepts become reality. Cadence's customers are the most innovative companies around the world, delivering everything from chips and circuit boards to the most dynamic application markets such as consumer electronics, hyperscale computing, 5G communications, automotive, mobile, aerospace, industrial and medical. Systematic excellence in electronics. Cadence has been ranked among Fortune magazine's 100 Best Companies to Work For for six consecutive years. For more information, please visit the company's website at cadence.com.
© 2021 Cadence Design Systems, Inc. All rights reserved. All rights reserved worldwide. Cadence, the Cadence logo and other Cadence marks listed at www.cadence.com/go/trademarks are trademarks or registered trademarks of Cadence Design Systems, Inc. All other marks are the property of their respective owners.
Recommended reading
京公网安备 11010802033920号