中文 English

Week In Review: Auto, Security, Pervasive Computing

ADAS fails in rain; Toyota’s $3.4B battery investment; safety SW; supply-chain security; swarm robots.

popularity

An investigation by the Automobile Association of America found that lane-keeping assist and automatic emergency braking, both high-profile ADAS features, are prone to failure in rain. According to the report, 69% of tests conducted with simulated rainfall resulted in test vehicles crossing lane markers, and 33% of simulations resulted in collisions at 35 mph. Surprisingly, risk of accidents did not increase with a dirty windshield.

Apple rolled out a new version of its high-end MacBook Pro that includes a complex, heterogeneous architecture designed for optimal data throughput. The new M1 Max SoC includes a 10-core Arm-based CPU, up to 32 GPU cores, a 16-core neural engine, and a media engine. Memory bandwidth is 400 GB/s, which is twice the throughput of its base-model M1 chip. Apple previously relied on Intel chips for its high-end notebooks, but as a vertically integrated company it has the advantage of being able to customize chips for software, and vice versa, greatly improving performance for specific applications.

Automotive
Toyota will invest $3.4 billion in auto battery development and production in the United States over the next 10 years. That investment is part of the total $13.5 billion Toyota will spend worldwide on battery technology.

Cadence introduced a Safety Solution platform targeted at safety-critical applications, including analog and digital safety flows, and tools for faster ISO 26262 and IEC 61508 certification.

The Netherlands Forensic Institute decrypted Tesla‘s driving data, which includes everything from driver assistance systems to accelerator pedal position, steering wheel angle and brake actuation. NFI said now that it knows the kind of data being collected, it can ask for specific data related to accidents. That kind of data is essential in a crash.

Coincidentally, the U.S. National Transportation Safety Board this week updated its investigation into the April 17 fatal crash of a Tesla Model S in Texas. Investigators found the driver had his foot on the accelerator pedal leading up to the crash, and that it was depressed as much as 98.8%. At its peak, the car was traveling at 67 mph in a 30 mph zone. The car drove over the curb, hit a drainage culvert, a manhole, and a tree, at which point the lithium-ion batteries caught fire (see below).


Diagram of reconstructed Tesla crash in Spring, Texas. Source: NTSB

Deals
Synopsys is working with TSMC on new tools for 3D packages that can simplify integration of multiple components. The effort initially will be aimed at the high-performance computing market, which is where many of these packages initially are headed due to the cost and complexity, but they almost certainly will be extended to other applications as the technology becomes more mainstream.

Arteris IP inked a deal with Eyenix, which licensed Arteris’ FlexNoC Interconnect IP for high-resolution imaging in the security and surveillance market.

Infineon is working with Picovoice to bring voice AI to IoT and other edge devices based on Infineon’s PSoC 6 microcontrollers. The speech recognition technology leverages an ultra-low-power architecture with a MEMS microphone.

Siemens said that Arm accelerated validation of its IP by 1,000 times using machine-learning software as compared to brute-force approaches.

Security

Microsoft, Intel, and Goldman Sachs are teaming up in what they’re calling a Trusted Computing Group. The goal is to improve supply chain security.

Takayuki Kobayashi, Japan’s new minister for economic security, said his country needs a strategic vision that matches approaches taken by the U.S. and China, without which it risks being marginalized. Kobayashi also noted that Japan’s relationship with China is extremely important, both for Japan as well as for regional and global communities.

The U.S. Cybersecurity & Infrastructure Security Agency issued an alert about BlackMatter ransomware, which so far has targeted critical infrastructure entitites, including two from the U.S. Food and Agriculture sector. The malware targets the Lightweight Directory Access Protocol (LDAP) and Server Message Block protocol to access the Active Directory, gaining information about all hosts on the network. It then remotely encrypts those hosts and shared drives.

The ISA Global Security Alliance and ISA Security Compliance Institute rolled out a product certification study based on the ISA/IEC 62443 standard, which seeks to manage security risks for industrial communication networks.

Robotics
If your restaurant meal doesn’t arrive quickly enough, you may have to blame the technology. Reeling from labor shortages, restaurants have begun using robots to deliver food to tables. That should create some interesting conversations about tipping.

Researchers at  University of Notre Dame have developed four-legged robots that can work together as a swarm, overcoming problems together that a single robot cannot. The researchers noted that more sensors are needed, and battery technology must be improved, but the ability to leverage sensors from multiple robots rather than just one has broad implications for military, scientific, and commercial applications.

Sword Defense Systems has created an unmanned rifle pod that fits on a dog-like robot and can shoot with machine-guided precision. The rifle pod can be loaded and unloaded remotely by an operator for safety reasons.

Certifications
Synopsys’ Fusion and Custom Design Platforms were certified for TSMC‘s N3 and N4 processes. The tools provide faster timing closure and full-flow correlation, from synthesis to timing and physical signoff.

Cadence‘s digital and custom analog flows also were certified for TSMC’s N3 and N4 processes, and are targed for mobile, AI and hyperscale computing.
Cadence also demonstrated IP test silicon for PCIe 6.0 on TSMC’s N5 process. Cadence said design kits are now available for early adopters.

And Finally
Inserting AI and machine learning into chips adds a whole new dimension of complexity, and creates a variety of potential problems, including deadlocks, loss of performance, and difficulty in achieving closure on many fronts in this recent Semiconductor Engineering video.

Find Semiconductor Engineering’s latest Auto, Security and Pervasive Computing newsletter here.

Get a sneak peek at the Technical Paper page Semiconductor Engineering is building. We’re in the process of gathering content for a Technical Papers page for all technologies related to the chip industry. If you have a recent published paper that is relevant to the Semiconductor Engineering audience and would like it to be highlighted on the site, please send a link to linda@semi-mags.com. SE does not typically publish entire papers, but we do link to proper sources. This page will not include company-specific white papers.



Leave a Reply


(Note: This name will be displayed publicly)