Autonomous vehicles (AVs) may well be the quintessential symbol of our AI/robotics age. Cars are a major part of our daily lives. A self-driven car is a concept that requires little explanation, something we can all quickly grasp. It wasn’t that long ago that the notion on an AV driving us around while we read or sleep would have felt like the stuff of science fiction. Having experimental AVs coursing through public roads in Silicon Valley, Pittsburgh and Phoenix is concrete evidence that our smart machines are achieving human-like intelligence, raising a number of important questions: how long before AVs are all around us?; how will they impact our lives?; what unintended consequences might we have to deal with?; and what should be done to ensure that they arrive as safely and smoothly as possible?
These are among the questions addressed in the March 1 issue of The Economist, which includes a comprehensive special report on autonomous vehicles with seven articles on the subject. The special report starts out with the assumption that whatever technological hurdles lie in their way will be eventually overcome. But there are wider economic, social and public policy issues to be explored, starting with: what can we learn from the transition to horseless carriages in the 20th century that can now be applied to the transition to driverless cars?
At the turn of the 20th century, all big cities were grappling with the growing volumes of horse manure and the diseases being spread by thousands of dead horses. By comparison, cars seemed clean and hygienic, and promised to provide safe and fast transportation, - key reasons why they were so quickly embraced.
Cars granted us enormous personal freedom and changed the world in all sorts of unforeseen ways. But in return, they were accompanied by unintended consequences and heavy societal costs. Just like cars were first viewed as a fix for the problems caused by horses, people are now looking to AVs to help us address the problems brought about by cars, especially accidents, pollution, and congestion.
Accidents. The National Highway Traffic Safety Administration (NHTSA) estimates that 37,461 lives were lost on U.S. roads in 2016, an increase of 5.6 percent from calendar year 2015. In addition, over 2.4 million people were injured in 2015. 94 percent of serious crashes can be linked back to human choices and errors. The Economist notes that around 1.25 millions people die, and another 20 - 50 million are injured in road accidents around the world each year. It’s the leading cause of death among young people aged 15-29.
Autonomous vehicles promise to drastically reduce these numbers. Even at this early stage, there’s evidence that AVs are considerably safer than human-driven cars, - not surprisingly given the high percentage of crashes due to human error. Moreover, safety is the highest design priority for AVs, because to succeed they will have to be almost infallible. While people tolerate deaths caused by human drivers, they’re much less likely to do so when there are no humans involved.
Pollution. Electric cars are much cheaper to run, so most AVs will almost certainly be electric, especially those used as robotaxis whose high utilization makes it imperative to have low operating and maintenance costs. This should help reduce harmful emissions, particularly in high density urban areas. In addition, electric vehicles are much quieter, helping to reduce noise pollution.
Congestion. It’s less clear how AVs will help reduce congestion, and it will likely take time. Computer-controlled AVs should be able to optimize their overall route planning to ease congestion. Over time, they should be able to travel closer together than human-driven cars, thus increasing road capacity.
Autonomous Vehicle Technologies
In just a few short years, AV technologies have advanced from not-sure-it-can-be done to it’s-just-a-matter-of-time. Google, Uber, and just about all auto companies are now developing autonomous vehicles. But, there’s still a way to go before the technology is ready for mass deployment.
“A fully autonomous car must solve three separate tasks: perception (figuring out what is going on in the world), prediction (determining what will happen next) and driving policy (taking the appropriate action),” explains The Economist.
Perception. AVs use a combination of technologies to perceive the world around them, including cameras, radar and LIDAR, - a method that creates a high resolution 3D map by using pulsed laser light and measuring the reflected pulses with sensors. These various techniques have different strengths and weaknesses that complement each other. Cameras are cheap, for example, but cannot measure distance, while LIDAR provides fine detail including distance but is expensive.
Prediction. To predict how the objects around it will behave, the AV needs to first identify what those objects are - other vehicles, pedestrians, cyclists, road signs, and so on. The hardest things to identify are unanticipated objects such as debris, roadwork, broken-down vehicles, and accidents. Rain, snow and puddles can also confuse an AV. Once everything has been identified, the AV has to predict how they’ll behave in the next few seconds and then determine what actions to take.
Driving actions. AVs are at a considerable disadvantage compared to human drivers, who’re used to dealing with exceptions to the normal flow of traffic. In the foreseeable future, - even as technologies continue to advance and AVs are widely deployed, - it’s quite likely that AVs will need to ask for human assistance every so often. The humans providing the assistance will likely be in central control rooms, will have access to all the data the vehicle is receiving, and will be able to direct the vehicle as to what it should do or even to take control and remotely guide it to get around problems.
Advances in machine learning will be a great help, especially as AVs can learn from each other’s data and experiences. Over time, there will likely be road lanes and entire areas dedicated to AVs, where vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) technologies will allow AVs to better coordinate their actions.
The Economist predicts that self-driving vehicles will be initially deployed by fleet operators, not by individual owners, for two main reasons. First are the high costs, not only of LIDAR and other technologies, but also of the required support structure. Those costs make business sense when amortized over a fleet of vehicles in use almost all the time but not for a private vehicle which is only in use 5% of the time. In addition, AVs may be initially limited to dedicated areas where they would be widely used, such as dense city centers, which is fine for a robotaxi but not for a private car.
“It is likely to be many years before AVs are cheap enough for individuals to buy them, and capable enough to operate outside predefined, geofenced areas. Meanwhile, the roll-out of cheap robotaxis in urban areas might encourage many young urbanites, who are already going off car ownership anyway, to abandon it altogether.”
The advent of AVs will have a profound impact on the auto industry. Car markers will have to reinvent their business models. “In an autonomous future where ownership is optional, they need to be selling rides, not cars.” Such a shift could open new opportunities. Whereas the car market is worth around $2 trillion a year, the market for personal transport is significantly bigger, perhaps as much as $10 trillion a year. That’s why it’s attracting a number of new competitors, including technology companies like Google and ride-hailing services like Uber.
People who drive taxis, delivery vehicles and trucks are clearly threatened by AVs. But it’s quite possible that their jobs will be redefined rather than abolished. “Delivery drivers could be employed to manhandle large packages into customers’ homes. Truck drivers might become overseers of platoons of vehicles traveling on highways. And AVs will create new jobs for remote fleet supervisors and mobile repair workers.”
As has been the case with cars, AVs will have a major impact in our personal lives, in communities and in society in general.
Cities and their associated metropolitan areas will be transformed in ways that we don’t yet understand. “Urban freeways, commuter suburbs and mandatory parking requirements reshaped cities. Now AVs promise to transform them once again, undermining many car-centric assumptions made in the 20th century, opening up new possibilities and turning urban-planning debates upside down.”
Regulating such complex, impactful, rapidly evolving technologies is very hard. Policymakers must ensure that AVs arrive safely and smoothly without inhibiting innovation. They should work closely with AV firms to develop new safety standards, issue guidelines, and permit limited testing on public roads. But they should wait for evidence that the vehicles are safe before approving widespread deployment.
The process has been compared to that of developing a new drug. “First you show in the laboratory that it might work; then you run clinical trials in which you carefully test its safety and efficacy in real patients; and if they are successful, you ask for regulatory approval to make the drug generally available. On this analogy, autonomous cars are currently at the clinical-trial stage, without final approval as yet.”
“A century ago cars raised fundamental questions about personal autonomy, freedom of choice and mobility,” said The Economist in conclusion. “AVs will do the same again. But this time around, with the benefit of hindsight, there is a chance that they will be seen not simply as a new form of transport but as a technology with far-reaching social and economic implications. Driverless cars present an opportunity to forge a new and better trade-off between personal mobility and societal impact. But AVs will deliver on their promise only if policymakers - like passengers climbing into a robotaxi - are absolutely clear about where they want to end up.”