Testing: The Key to AI-Enabled Autonomous Driving

By Emmy Yi, SEMI Taiwan

Since 2010, 474 companies have poured $51 billion into developing products enabled by artificial intelligence (AI), with the bulk of these investments targeting autonomous driving and in-vehicle experiences, according to the McKinsey reports. With AI and automotive electronics promising massive growth potential, it’s no surprise that IHS Market predicts the Advance Driver Assistance Systems (ADAS) market will reach $67.43 billion by 2025 and that, by 2040, 33 million AI-enabled autonomous vehicles will be on the road worldwide.

Industrial experts from MediaTek, TSMC, KYEC, Advantest, YTEC, and Cadence share their perspectives on testing.Lured by this immense business opportunity, many key semiconductor industry players are jumping into the automotive market. Their ICs, however, will face far more stringent reliability requirements than most consumer products, making testing crucial to accelerating the realization of level 5 autonomous driving – a fully autonomous system that rivals the behind-the-wheel performance of a human driver even in extreme road conditions like snow and ice.

With testing such a vital aspect of autonomous driving, SEMI Taiwan recently connected industry experts from IC design and testing-related fields to facilitate cross-discipline collaboration and help inspire innovative solutions to current testing challenges. The early February AI IC and Automotive IC Test Seminar is part of a series of SEMI Taiwan events focused on hot topics including like AI, IoT, smart automotive, smart data and smart MedTech. Following are a few key takeaways from the seminar.

A Paradigm Shift is Needed in Automotive Electronics Testing Strategies 

Adding AI to already electronically complex automotive systems will dramatically increase the number of ICs and sensors in vehicles. Designers of automotive electronics need to transform their test strategies to match the technical rigors of autonomous driving. The traditional process of build, test, and then fix-for-compliance must be changed in the era of self-driving vehicles. Adding AI to already electronically complex automotive systems will dramatically increase the number of ICs and sensors in vehicles.

The process of testing component points of failure falls well short of the requirement to test under countless driving scenarios during a which a device might fail. Testing, therefore, must be holistic. Starting in the development phase of their own electronics systems, automotive electronics designers must work closely with component and other technology suppliers to ensure that designs are well-integrated and rigorously tested for interoperability and points of failure under any conditions a human driver would face.

Wafer-level Test Is a Trend

The cost and time for IC testing have steadily increased to meet the relentless scaling requirements of highly integrated advanced technologies, placing immense pressure on current wafer-level packaging and testing methodologies to maintain cost efficiencies, chip yields and time-to-market speed. The challenges will intensify with the multiple-component parallel testing required for autonomous vehicles. Demands on automotive electronics manufacturers to maintain DDPM quality levels key to smart functionalities, powertrain operation, safety and reliability will also complicate current IC testing methodologies.

To fulfill the promise of autonomous automobiles and other AI applications, industry, academia, and government in Taiwan must work together to solve underlying technical challenges, create profitable business models and develop a strong programming and system integration workforce. Taiwan has a solid foundation to build on. With its strong semiconductor manufacturing industry and advanced IC testing capabilities, Taiwan is well-positioned as a growth engine for the advanced automotive electronics needed for autonomous vehicles.


x Brown