The Passenger Economy: Challenges Ahead
July 31st, 2017
By Lynnette Reese, Editor-in-Chief, Embedded Intel Solutions and Embedded Systems Engineering
As horseless cars become driverless cars, plan to see a tremendous increase in productivity in every area that transportation touches today, and in applications yet to be imagined.
Early Autonomous Vehicles (AVs) are here but pose grave engineering challenges if we are to implement them in real life. We are witnessing a moment of unprecedented transformation in automotive history. The time when AVs are as common as smartphones is a couple of decades or more away, but the potential for new markets and significant increases in productivity are seductive.
Autonomous Vehicles to Save $1.3 Trillion a Year
Morgan Stanley predicts that the U.S. alone stands to save $1.3 trillion a year, presumably with a full-scale implementation of AVs. Nearly $500 billion annually would be saved in accidents due to reductions in repairs, medical bills, and lost time, with another $500 billion in overall productivity gains. Improvements include improved logistics, lower fuel consumption, less traffic congestion, and personal productivity gains (assuming the average person drives an hour and a half each day). Time spent commuting, finding, and paying for parking in crowded cities will be significantly reduced as everything runs more efficiently. Sharing vehicles means fewer cars on the road and a lower cost of living for individuals while maintaining autonomy. The vision for a world with AVs includes less traffic, better maps through detailed communications from intelligent cars, and mobility for all. A recently released report by Strategy Analytics includes some tantalizing predictions, such as lower public safety costs due to traffic accidents that could total over $234 billion from 2035-2045. Commuting time saved by autonomous vehicles may amount to over 250 million hours per year for consumers in the most congested cities1.
Once people cease to drive, what do they do? A new market will open up, called the “Passenger Economy,” a term coined by Intel® CEO Brian Krzanich. Different applications, markets, and businesses can shoot out of this inflection point in how we carry out our lives. Consumer services such as entertainment, advertising, and personal or financial services might be carried out inside autonomous vehicles as people travel to work. Intel’s Katherine (Kathy) Winter, Vice President and General Manager of the Automated Driving Division, deliberates a passenger economy and the new markets it will incite. “It’s probably the smaller piece right now, but mostly because we can’t imagine what it is yet,” she states. Winter suggests that this portion of the passenger economy would include “the new services that a person in an [autonomous] vehicle could be consuming; something [like] entertainment, education, advertising, things like that.”
As a population of drivers evolves into passengers, Intel, working with Strategy Analytics, finds that by 2050 the passenger economy is an estimated $7 trillion industry.1 Winter indicates that the semiconductor industry must squeeze yet more improvement in computing power to meet technical challenges that autonomous driving poses, including leaving “headroom” for things we cannot imagine yet. The largest factor holding the passenger economy back will likely be social, not technical, as new laws and regulations must be crafted. Consumer acceptance may be slower than technologists and early adopters would like, as trust builds in AVs and jobs are re-established in areas related less to the actual driving and more to logistics management.
What can we expect? Winter points out the trucking industry as one example where Mobility-as-a-Service (MaaS) will dominate. “Look at the trucking companies. It’s well known there is a shortage of truckers, people wanting to drive long-haul trailers, and the need to move packages around is escalating…business [can take] advantage of Mobility-as-a-Service in not having to have drivers. This is time and money for them.” The technical challenges to attaining AVs are not small. AVs will shift the semiconductor industry’s focus to processing teraflops of data at blinding speeds to fulfill the need for low latency, high bandwidth, and rapid throughput on a scale that includes data creation and consumption the likes of which we haven’t yet seen as commonplace.
The Technical Trinity: Car, Cloud, and Connectivity
The picture for AVs includes the car, a cloud, and connectivity. AVs are loaded with sensors to create enough information on which to make split second decisions necessary to drive. Myriad sensors are loaded onto AVs, including but not limited to radar, LIDAR, cameras, infrared sensors, and GPS. Many of these sensors are placed all the way around the vehicle and work together to build the car’s vision of what’s going on around it and where should it go next. Varying environmental conditions make redundancy and overlapping data input necessary to fuse together a knowledge that is more reliable than mere vision as to what is going on around, behind, and ahead of the car so that decisions can be anticipated. Changing environmental conditions include road conditions, variations in street layouts, random or sudden incidents involving bicycles, pedestrians, or animals, and changing weather conditions that can affect some sensor’s ability to gather accurate data more than others. There’s a reason why Phoenix, Arizona is such a popular testing ground for AVs these days. Winter points out that Phoenix is laid out like a grid and that testing engineers are pretty much counting on it being sunny. “You’re not going to see snow, you’re not going to see ice. But we need vehicles that can drive autonomously every day of the year, not just when it’s sunny, not just when it’s not raining or snowing, or below 32 degrees.”
AV prototypes must test to optimize the right mix of sensors for different environmental conditions. Thus, accurate sensing in every kind of condition is part of the challenge to making the passenger economy a reality. Why do we need all these sensors? “They all complement each other, which is kind of the bottom line. And as we go through more and more testing, and there’s more of those vehicles out there, we are learning about the combinations, how much redundancy, things like that, that you actually need in the vehicle.”
Proving Grounds for High-Performance Computing
High performance embedded computing (HPEC) is another area that challenges the trinity of car/cloud/connectivity also known as “the three Cs.” Extremely fast, low latency, high-throughput devices that are extremely secure are on the agenda for acceptance and success of this pending phenomenon. Winter comments that today, the estimated average amount of data that a person generates in a typical day is around 650 MB today. By 2025, this figure might be nearer to 1.5 GB per day. However, AVs produce an estimated 4,000 GB per day. How do you get to that number at scale? Winter states that an average vehicle that’s driven an hour and half a day yields “that kind of data in the vehicle itself.” This translates to a challenge for HPEC. “Today in 2017, there’s probably half a teraflop of data coming off those [self-driving] cars, on an average day. If you go out to 2025, you think about all the sensors, all the data, everything coming off there, look at the computing power needed to process, that, store, share, save, et cetera.”
Other serious challenges include where and how to rapidly store and retrieve enormous amounts of data, and how to manage, process, save, store, and share that data from each and every car on a scale that’s beyond anything we have seen to date for an infrastructure that stretches coast-to-coast. Each state has differing traffic laws, landscape, and weather. A large number of servers positioned across the nation will handle storing data that’s needed to perform deep-learning processes, and create algorithms for use by a massive fleet of AVs that will learn collectively from what can only be described as each other’s learning experiences. And not all data is equal. How do we decide what is safety-critical? If safety-critical data is obtained, how do we get it to a fleet of cars on a massive scale, and as soon as possible? There’s lots of learning and testing to do, and as yet, communications technology is missing a critical piece.
Connectivity
Communication is the last vital piece to work out in how the trinity of car, cloud, and connectivity brings AVs to full scale. Wireless 5G communication has been on the radar for some time now as a well-needed means of upgrading cell phone communications to much faster, higher bandwidth wireless services. Intel has long recognized 5G as ideal for industrial use cases, smart cities, and anything that requires super high bandwidth, super low latency, and wide-ranging wireless connectivity. But Intel has found that the use case for AVs has taken the lead for 5G, according to Winter. The autonomous vehicle “seems to be one of the forces really driving the need for [5G], and the lead use case from what we can see. You need a way [to communicate], and we’ve talked about the end-to-end, mission critical information, that needs to be really fast, low latency [communication]. The other piece is the pure volume of all that data.” Once 5G is out there, cars can use it to communicate to one another, to the cloud for critical updates both ways, and to and from surrounding infrastructure. It may be that speed limit signs will disappear one day, replaced by innocuous transmitters that signal the speed limit to passing cars. Stoplights may transmit status to vehicles directly. The anticipated 5G also adds another sensor to the car.
So How Do We Get There?
Labor costs may go down with AVs, but we will see an increase in demand for cloud services, servers, predictive analytics, and Internet of Things (IoT). Data collection, processing, storage, retrieval, and analytics will be even more important as we analyze, predict, optimize, decide and anticipate using high-performance computing as just part of the picture. At a full-scale implementation, the challenges of a driverless world seem tremendous. One step at a time, building on the work of others, is how science and technology have evolved for centuries. What Intel has been pushing for, and what Winter believes is critical to accelerating success, is in standards and collaboration across the industry. Winter believes that success is in developing talent for the task and in “sharing potentially mission-critical safety data…across the industry, so that every vehicle manufacturer, every fleet operator, doesn’t need to put in millions and millions of miles to understand the safety aspects of driving autonomous vehicles. Sharing that kind of safety data will accelerate the pace of the industry.” Additionally, shared standards mean that no one has to learn and develop uniquely, on their own. “Putting some standards and industry-wide platforms in place will help accelerate the entire industry. If you really think about this from a safety perspective…it’s worth doing for everybody.”
Winter says that one of the most common questions she hears is “What will it take to make this happen?” Winter states, “To make this happen, the biggest thing we need to do is trust and just let go. Let go of the wheel.”
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.
1. “Intel Predicts Autonomous Driving Will Spur New ‘Passenger Economy’ Worth $7 Trillion.” Intel Newsroom. Intel Corporation, 1 June 2017. Web. 26 July 2017.