Intel GO Solutions Pave the Way for Autonomous Cars
If the destination is a place where solutions can take a big bite out of transportation costs…are we there yet?
In 1978, Cadillac introduced a trip computer, “a device set in the dashboard that is equivalent to a programmable calculator in the home. The computer can work out…a driver’s fuel consumption or the number of miles before he reaches his destination.” Car makers envisioned a driver information center that would extend dashboard controls, front radar scanners to determine a safe distance from the car in front, and keyless entry (Marsh, 1979). Forty years later, we are promised self-driving vehicles by 2021, if Brian Krzanich is correct in estimating that this is when self-driving vehicles will be out and about with the Intel® GO™ system aboard.
Intel is correct to delve into automotive relationships concerning assisted and autonomous driving applications, since Level 5 autonomous cars are the holy grail not only for tech companies, but for agencies such as the National Highway Traffic Safety Administration (NHTSA). In 2014, the Society of Automotive Engineers (SAE) organization published a standard (J3016) for defining five levels of vehicle automation, from Level 0 at no automation through Level 5 at fully autonomous, with no human intervention. Level 1 is described as driver assistance, which might include cruise control, for example. Level 2 includes partial automation, where the system will execute aspects of both steering and acceleration and deceleration, for which adaptive cruise control qualifies. Level 3 is conditional automation, where the system not only executes Level 2 automation, but also monitors the driving environment. With Level 3 comes an expectation that the human driver will intervene if prompted to do so. One might consider Tesla’s autopilot as a Level 3 task. Level 4 covers some modes of driving and continues all aspects of Level 3 but does not require driver intervention for that driving task. This may include a future driving mode for highway-only driving but not in hazardous weather, for example. Level 5 is fully autonomous driving where all aspects of driving that humans would perform are done by the autonomous vehicle. No human intervention is expected at Level 5 (see Table 1).
Safety Is Just One Advantage
In 2016 there were 261.8 million registered cars and light trucks in the United States. According to former NHTSA Administrator Mark Rosekind, 94% of all vehicle accidents are due to human choice or error. Highly Automated Vehicles (HAVs) have become a focal point for increasing safety in transportation. According to the Federal Automated Vehicles Policy, “While a human driver may repeat the same mistakes as millions before them, an HAV can benefit from the data and experience drawn from thousands of other vehicles on the road.” The data from millions of automated and learning vehicles on the road will add safety to the list of benefits of big data. However, the NHTSA will “…continue to exercise its available regulatory authority over HAVs using its existing regulatory tools: interpretations, exemptions, notice-and-comment rulemaking, and defects and enforcement authority.” Autonomous Vehicles (AVs) are a serious business and much more relevant than past innovations of convenience such as Cadillac’s trip computer.
So, it’s no wonder that fully automated driving is attractive; companies are racing to get autonomous cars on the road. Safety is just one advantage, however. Autonomous vehicles can ultimately reduce not only accidents, but traffic congestion, air pollution, and fuel consumption while reducing the need to expand highway infrastructure as self-driving cars change the American paradigm for mobility with Uber-like cost models for car ownership and use. Self-driving cars also offer independence to the disabled and the elderly. At a critical mass/point juncture, we will see a tipping point where AVs are the majority of the vehicles on the road with a potential for significantly bringing down the total cost of transportation. That tipping point, however, where a total number of AVs brings down the cost is a long way away; Rosekind also stated that older vehicles will remain in regular use on the roads an additional two or three decades after AVs come into serious use. Nevertheless, Rosekind’s two-year tenure hallmarked the cooperation of the automotive industry in agreeing to make automatic emergency braking standard equipment by 2020, as well as a Federal Automated Vehicles Policy to “…ensure these technologies are safely introduced (i.e., do not introduce significant new safety risks), provide safety benefits today, and achieve their full safety potential in the future.” Cars are safer than they used to be, but texting while driving can be added to the catalog that has listed driving under the influence for decades. Of late, technology that encourages driver distraction has increased recent highway death tolls in spite of technology-induced safety improvements.
The five largest chip makers currently serving the automotive market are NXP, Infineon, Renesas, STMicro, and Texas Instruments. Intel wants to change that. At the 2017 Consumer Electronics Show (CES), Intel, BMW AG, and Mobileye announced at a joint press conference that their seven-month old partnership will soon produce 40 autonomous vehicle cars for testing on roadways by the second half of 2017. The companies have “developed a scalable architecture that can be adopted by other automotive developers and carmakers … from individual key integrated modules to a complete end-to-end solution providing a wide range of differentiated consumer experiences.” Mobileye, founded in 1999, is a leading supplier for core SoCs that go into vehicles with Advanced Collision Avoidance Systems. Mobileye’s EyeQ® chip technology supports features such as vehicle and pedestrian “up ahead” warnings to support collision avoidance, all with a single camera. The partnership will extend Mobileye expertise to “the development of fusion algorithms … deployed on Intel computing platforms.” Intel’s computing power scales with solutions that include the Intel Atom™ or Intel Xeon® processors “with up to a total of 100 teraflops of power efficient performance without having to rewrite code.”
Intel’s plans include not only high-performance computing for driver assistance to the autonomous car, but 5th generation cellular (5G) wireless connectivity and cloud. Intel GO Automated Driving solutions were introduced at CES 2017. The Intel GO In-Vehicle Development Platform for Automated Driving comes in an Intel Xeon version and an Intel Atom version, both with Intel Arria® 10 FPGAs for parallel processing. (Recall that Intel acquired FPGA-maker Altera in late 2015.) These boards provide a rapid and reliable way to develop, implement, test, and optimize everything from Advanced Driver Assistance Systems (ADAS) all the way through Level 5 fully automated driving without having to design hardware from scratch. The Intel Xeon version of the Intel GO platform enables solutions all the way to autonomous vehicles. The Intel GO automotive software development kit (SDK) allows developers to access the system for faster time-to-market with tools that incorporate computer vision and deep learning tool kits so they can develop and optimize algorithms for detection, sensor fusion, and execute on decisions. Sample reference applications for lane change assistance and object avoidance shorten the learning curve and time to market.
In addition to high-performance computing, Intel GO systems include software development tools, 5G-ready connectivity, a robust data center platform, and the latest in Artificial Intelligence (AI.) Automotive Intel Xeon processors, Intel Atom processors, and Intel Arria 10 FPGAs form the foundational basis for Intel’s vision of autonomous vehicles. The included Intel Arria 10 FPGAs “feature hard floating-point digital signal processing (DSP) with speeds up to 1,500 giga floating-point operations per second (GFLOPS).” The Intel Atom and Intel Xeon-based GO platforms (with Arria 10 FPGAs) come with sample applications, run time libraries, and middleware.
Intel supports connectivity in automated driving with the newly announced 5th generation (5G) cellular modem slated for release in the second half of 2017. 5G cellular communication is not going to be wide-spread until after 2020. Over-the-air (OTA) automotive updates allow car owners to avoid going to the dealership for updates. But 5G is more than just a channel for timely updates; it’s a means for intelligent cars to communicate with other intelligent cars, with surrounding infrastructure such as smart “signs” with information on detours, speed limits, and warnings, and with pedestrians for a variety of reasons. Smart cars can communicate with a “smart” city to reduce traffic congestion and facilitate large gatherings such as conferences, marathons and festivals that require street closings. Presently, we can flip a turn signal to indicate intention, but 5G can be used to communicate a sudden need to swerve to avoid an obstacle. Thus, an intelligent car can alert other cars via 5G, as well as report roadway hazards to the local department of transportation. Not all 5G advocates feel that 5G is necessary for autonomous driving. However, the 5G Infrastructure Public Private Partnership (the 5G PPP), a European association that was initiated by the EU Commission, the telecommunications industry, small and medium enterprises, and researchers, believes that autonomous vehicles are not safe without 5G communication. Existing LTE cellular systems have latency that would negate a portion of the safety aspect that is gained with wireless communications to and from autonomous vehicles. To effectively support autonomous driving, wireless communication will need to meet minimum metrics for latency, reliability, throughput in heavy network traffic, and coverage.
5G will also allow fleet managers to monitor their fleet more closely, with more accurate knowledge of the location of vehicles and goods in a fleet of trucks, for instance. Fleet management with automation means having more accurate estimates for shipping arrival times, improved overall asset management, predictive maintenance and accurate maintenance records, and eventually eliminate the trucker’s traditional role behind the wheel. Autonomous trucks will someday provide highly efficient driving control with lower instances of hard braking, softer starts, and constant attention to driving that will rival and surpass human driver capabilities. More efficient use of fuel and longer driving hours equate to improved profits. Autonomous cars will also make bus, taxi, and Uber drivers obsolete as scheduled and on-demand passenger pick-up and drop-off are usurped by self-driving vehicle services.
Lynnette Reese is Editor-in-Chief, Embedded Intel Solutions and Embedded Systems Engineering, and has been working in various roles as an electrical engineer for over two decades. She is interested in open source software and hardware, the maker movement, and in increasing the number of women working in STEM so she has a greater chance of talking about something other than football at the water cooler.
- Marsh, P. (1979, December 6). The making of the computerised car. New Scientist, 770-773.
- “Federal Automated Vehicles Policy.” Department of Transportation. N.p., Sept. 2016. Web. 27 Jan. 2017.
- WILMOT, STEPHEN . “How to Place Your Chips on Electric Cars.” Wall Street Journal 19 Jan. 2017: n. pag. Print.
- “BMW Group, Intel and Mobileye Will Have Autonomous Test Vehicles on the Roads by the Second Half of 2017.” Intel Newsroom. N.p., 4 Jan. 2017. Web. 27 Jan. 2017.
- “BMW Group, Intel and Mobileye Team Up to Bring Fully Autonomous Driving to Streets by 2021.” Intel Newsroom. N.p., n.d. Web. 27 Jan. 2017.
- Ma, Hui Sheng, Erqing Zhang, Shufang Li, Zhengnan Lv, and Jing Hu. “A V2X Design for 5G Network Based on Requirements of Autonomous Driving.” SAE Technical Paper Series (2016): n. pag. Web. 27 Jan. 2017.