Kevork Kechichian, senior engineering VP at Qualcomm Technologies Inc., delivered a keynote address on the future of automotive Friday at the 2nd IEEE International Workshop on Automotive Reliability & Test (ART17), held in conjunction with ITC. He described the car of the future as connected, self-driving, highly personalized, and efficient. Kechichian said he has spent his entire career in semiconductors, and although his background is not in test, he has managed test groups. He said he was looking forward to Qualcomm’s planned acquisition of NXP and the resulting potential to push further into automotive products.
As it happens, Kechichian was speaking one business day before Broadcom proposed to acquire Qualcomm Inc. (of which Qualcomm Technologies is a wholly owned subsidiary) in transaction valued at $130 billion. Broadcom said its proposal would remain in effect whether or not Qualcomm’s acquisition of NXP is completed.
But as the acquisition process plays out over the coming year, you can expect Qualcomm engineers to continue pursuing new technologies—including those related to automotive applications, from self-parking cars to ADAS. Lots of the necessary elements are already in place, Kechichian said, having been in the works for long time.
Self-driving vehicles drive a lot of technology on the hardware side, the system side, and the software side, he said, and Qualcomm is engaged in all of them with innovations in areas including embedded processing, DSP, machine learning, sensors, and 5G. Qualcomm, he said, will leverage its expertise in sensors, smartphones, radar, cameras, and edge processing to help pave the road to autonomy. He described the connected car as essentially an IoT device, albeit on a larger scale than a Fitbit, for example, and one that can offer the personalized experience and comforts of your living room. 5G and Wi-Fi will go into different aspects of the car, he said. Processing will occur at the edge of the cloud, with local decisions happening in real time. And manufacturers are driving zero-defect rates for mission-critical applications, he said.
He described a smart automotive infrastructure aimed at collision control, traffic management, and security that will maximize efficiency (recommending the best routes, for example) and safety. He described cellular V2X, embracing vehicle-to-infrastructure (V2I), vehicle-to-vehicle (V2V), and vehicle-to-pedestrian (V2P). The V2V problem represents a statistical problem of multiple moving cars having to talk to each other, whether following or trailing each other but always needing to avoid collision. “The processing is amazing,” he said. “There is so much data going back and forth it can’t wait to go to cloud and back.” As for V2P, he described a pedestrian carrying a cellphone with the smarts to mimic human interactions we take for granted—eye contact between a driver and pedestrian about to enter a crosswalk, for example. “Putting all of that into an algorithm is one of the biggest challenges that we have,” he said.
Altogether we are contending with a huge IoT type of ecosystem, he said, citing forecasts of more than $2.4 trillion in economic output from the automotive ecosystem in 2035.
From Qualcomm’s perspective as a chipmaker, he said, challenges involve evaluating the relevant specifications and choosing the right hardware DSP functions and fat memory pipes to implement them, while contending with issues such as ELFR, redundancy, security, and signal integrity. He cited a need to control 11 displays in a single car. When first presented with that requirement, he said, he thought the specifier wasn’t serious. But between an instrument display, rearview mirrors, and backseat infotainment displays the count adds up quickly.
Paving the road to autonomy, he said, will require decisions regarding wireless charging, updating maps, and effective pedestrian detection—all algorithms suited to deep learning. Cloud-based neural nets can learn everything there is to learn—the challenge is to impart the necessary knowledge to a new car that can make its own decisions. How can a new car leverage that cloud-based knowledge?
As for Qualcomm’s performance in the automotive space so far, Kechichian said the company has more than 150 design wins across telematics, infotainment, and connectivity applications. Snapdragon automotive infotainment platforms have won new designs in 12 of the top 23 global automotive brands, he said. He also said Qualcomm Halo wireless EV charging technology has been successfully integrated and tested on number of vehicle platforms.
As for test and coverage improvements, he said, an important aspect is partnering with foundries on process technology. DFT and design teams must cooperate, he added, to improve at-speed test coverage and increase the functional vectors to cover what can’t be found with structural test. Product and test engineering must cooperate on outlier detection and DPAT. Another possibility is system-level test (SLT), as is design failure mode and effect analysis (DFMEA).
In conclusion, he said, “Test coverage is becoming very important.”
