Few technologies have been more anticipated heading into the 2020s than autonomous vehicles. Tantalizingly close and yet still perhaps decades from market adoption in some use cases, the technology is as promising as it is misunderstood.
You've heard the consumer hype, but what gets less ink are the transformative changes that autonomous vehicles will bring -- in some cases already are bringing -- to the enterprise.
Affecting sectors as disparate as shipping and logistics, energy, agriculture, transportation, construction, and infrastructure -- to name just a few -- it's hard to overstate the impact of the diverse and versatile set of technologies lumped into the decidedly broad category of 'autonomous vehicles'.
This guide will help you sort the hype from the business reality and tell you all you need to know about the autonomous vehicle revolution on the ground, in the air, and even at sea.
In 1939, General Motors predicted we'd have an autonomous vehicle highway system up and running by the dawn of the 1960s.
As with a lot of autonomous vehicle hype, that prediction was a tad premature, but it demonstrates the long history of autonomous vehicle development. By 1958, GM had created a prototype autonomous vehicle that was guided not by on-board electronics but metal indicators embedded in a roadway.
Modern iterations of autonomous mobility have concentrated on placing the burden of navigation with the brains of the vehicle, although efforts to create road infrastructure that communicates with vehicles are alive and well, and will likely play an important role in the future of autonomous road vehicles, beaming the latest traffic updates and alerts to vehicles on the road so they can route plan and react accordingly.
The first cars to use on-board LiDAR (Light Imaging Detection and Ranging) sensors to detect roadways were developed in labs in the 1980s, but it wasn't until the 1990s that autonomous vehicle development really took off. In 1994, two robotic vehicles, VaMP and Vita-2, developed by Daimler-Benz and autonomous driving pioneer Ernst Dickmanns drove semi-autonomously (some human intervention) more than 600 miles on a Paris highway in heavy traffic.
None of the team's competing in DARPA's first "Grand Challenge", which required participating teams to develop vehicles that could drive 150 miles unaided through the Mojave desert, took the $1 million prize, but five teams finished the following year.
Google began developing its first autonomous vehicle behind a curtain of secrecy in 2009, and major auto manufacturers were quick to follow suit. Out of those efforts, new generations of sensing, machine learning, artificial intelligence (AI), and communications technologies have given rise to autonomous vehicles used both on- and off-roads and highways, and have profoundly accelerated related technology sectors such as mobile robotics.
The technologies behind autonomous vehicles can be primarily broken down into three categories: sensors, control algorithms, and connectivity.
On the sensor front, LiDAR is one technology adopted by many (but not all) autonomous ground vehicle developers. The spinning lasers first caught public attention thanks to Waymo's early public tests of its autonomous vehicles, but the technology has been around far longer than that, dating back to the 1930s. LiDAR has been used in aviation, oceanography, archaeology, and other fields for decades, but the speciality nature of the applications has kept sensor prices high.
That changed practically overnight with the entry of new manufacturers and the birth of compact MEMS-based LiDAR, currently being tested by BMW and other OEMs, which was all prompted by the rush of development in self-driving vehicles. At its core, LiDAR uses lasers to measure distance to objects at the speed of light, useful for making quick work of mapping environments for a moving vehicle to navigate. But LiDAR comes with drawbacks, including limited range, shaky poor weather performance, and lots of moving components, which can cause reliability issues.
For some developers, the issues with LiDAR counsel investing in a variety of additional or replacement sensor technologies, including radar, which has exceptional range and is solid state, thermal cameras, which work well at night, and visual spectrum cameras (a favorite of aerial drone developers thanks to their low weight), and to advances in machine vision software. The fact is, no single sensor that exists today is perfect for all self-driving applications, but in concert a suite of sensors can provide excellent coverage in a variety of road, air, or undersea conditions. However, making it all work together in a way that's cost effective has been one of the persistent challenges in the space.
On the software front, advances in AI and machine learning are driving autonomous vehicle development. Machine learning has been especially important. The fact is, no engineer can program a vehicle to make the right choice in every scenario, but using machine learning protocols, vehicles can be taught to make the right decision through arduous road testing.
Then again, road testing has limitations. For one, it takes a long time to drive millions of miles. It's also not practical to expect road tests to account for every eventuality a vehicle is likely to face. High-fidelity physics simulation programs from companies like ANSYS, which enables OEMs to test all the sensor elements in real world conditions at scale without requiring millions of miles on physical roads, have provided an important stopgap and are speeding up development.
Connectivity is often overlooked in the self-driving vehicle conversation, but enabling vehicles to communicate with infrastructure, such as roads, and with each other is essential to creating safe self-driving architecture. Connected vehicle testbeds are operational in various states in the US, including in Michigan, where the Department of Transportation has collaborated on a testbed that consists of a network of 50 roadside equipment (RSE) units installed along various segments of live interstate roadways, arterials, and signalized and un-signalized intersections in Novi, Michigan, outside Detroit.
Aerial drones may also benefit from connecting to terrestrial infrastructure, which could limit risk associated with navigational malfunctions and lost connections as the skyways open up to commercial UAV. Because connectivity is severely impeded underwater, and because oceans are the least developed ecosystems on the planet, it's more challenging for developers of autonomous aquatic vehicles to cope with connectivity issues.
Planes, trains, cars, trucks, boats, subs, bots, bikes … There aren't too many mobile machines that aren't the subject of ongoing autonomous vehicle development. To narrow the lens a bit, in this guide we'll look at self-driving cars and trucks, UAVs, and unmanned aquatic vehicles.
State of the technology & regulation
Are consumers ready for autonomous cars? According to a recent survey commissioned by ANSYS, an engineering simulation company, the answer is a resounding 'yes'.
On behalf of ANSYS, Atomik Research, an independent market research firm, executed an online survey of 22,041 adults aged 18+ in countries considered to be prime targets for autonomous vehicle technology. The firm found that seven out of 10 consumers believe autonomous cars drive better than humans or will surpass human abilities by 2029. Following large-scale reevaluation on the part of regulators and developers of the pace of AV adoption following high-profile accidents, the results of the survey should be encouraging.
The technological barriers are falling thanks to new sensors, such as 3D semantic cameras, which use a low-powered laser to detect the chemical composition of various objects radar arrays. This can complement or replace LiDAR, and even a brewing showdown within the $1.1 billion LiDAR sector between amplitude modulation and frequency modulation. (Nearly all LiDAR today is AM, but companies like Aeva are making a strong case for FM).
Regulatory hurdles -- the biggest being a lack of regulatory clarity -- are also beginning to fall, at least on the state level. In California, which has been extremely quick relative to other states to enact regulation governing self-driving cars and trucks. Several laws outline protocols for the testing and deployment of driverless cars. More than 50 companies currently hold permits to test on roads in California, each having gone through a credentialing procedure with the CA DMV, and robo-taxi services will begin in some municipalities within the next year.
On the federal level, passing legislation has been more challenging. An open rule-making inquiry by both the National Highway Traffic Safety Administration (NHTSA), which regulates cars, and the Federal Motor Carrier Safety Administration (FMCSA), which regulates commercial trucks, may result in relaxed restrictions on several physical requirements for vehicles operating on US highways, including specifically human control devices. Transportation Secretary Elaine Chao is supportive, but the AV START Act, cosponsored by Senators John Thune (RSD) and Gary Peters (D-MI), which sought to establish federal standards for autonomous vehicle regulation failed in the senate in late 2018.
You'd be hard-pressed to find a major automaker not currently pouring development money into autonomous features for their vehicles. As a result, much of the autonomy coming to roads in the near-term will not focus on L4 self-driving, but autonomous features, such as lane-center steering, adaptive cruise-control down to a stop, self-parking, auto lane-change, and, in a few models, full autopilot modes.
But the rush to automate may also spell the end of an era for carmakers, who have long relied on a bedrock principle in many Western countries: Every person should have a car.
Now, ride sharing is reducing the incentive for young city-dwellers to buy cars, and that will only increase as ride-share companies ditch the drivers and automate. Imagine a future in which cars rove traffic-optimized streets, allowing passengers to step in and out of cars in a flash to hop across town. The arrival of blockchain, connected cars, renewed interest in urban transportation systems in cities like Los Angeles, and, of course, autonomous vehicles are fundamentally changing our relationship to cars, and that's going to have a negative long-term impact on automakers who fail to adopt a suitable business strategy.
There seems to be a new autonomous delivery company hitting the road with L4 autonomy every week. One of them is Nuro. Founded by former Google engineers Jiajun Zhu and Dave Ferguson, Nuro is launching a small, self-driving delivery van for last-mile urban delivery.
Nuro's pitch to consumers (and, inevitably, to local and federal legislators still wary of self-driving technology) will be that its pint-sized, battery-powered vans are safer than human-piloted vehicles and will relieve congestion and pollution by taking bigger delivery trucks off crowded city streets.
Another company making strides with real-world tests is AutoX, which was founded by a young Princeton Professor named Jianxiong Xiao (Professor X to his students). AutoX is now delivering groceries via its L4 autonomous vehicles in San Jose, with additional public trials to follow.
Unlike most deep-pocketed self-driving companies, which use laser arrays and a bevy of sophisticated sensors in their autonomous platforms, AutoX uses inexpensive cameras and relies on its AI smarts to navigate the world around it (Tesla has also forsworn LiDAR in its Autopilot mode).
AutoX is partnering with GrubMarket.com for its new grocery delivery service. A temperature controlled AutoX vehicle arrives with groceries and pops the trunk. Unlike Nuro, which was founded by two former Google engineers, AutoX uses OEM consumer vehicles retrofitted with its technology.
These are just two of the many companies competing in a space that's about to blow up in a big way. It should also change the fates of smaller chains and even mom-and-pop shops, long bested in the delivery game by Amazon. Many of the autonomous delivery companies are heralding as-a-service models that should bring fast delivery of everything from groceries and meals to parcels to new sectors and business.
Part of the appeal of autonomous driving to enterprise customers will be that these vehicles save money by optimizing delivery routes, decreasing accidents, and reducing workforces.
One question that gets asked a lot is whether, at some point, it will be illegal to drive a non-autonomous vehicle on the street. It's possible, but I'd put my money on a market-based solution.
Namely, once a majority of the fleet on roads becomes autonomous, it will become prohibitively expensive for most would-be drivers to afford insurance for their 'classic' car in everyday use situations. In fact, it stands to reason that in an autonomous future there won't be more than a handful of niche insurers for piloted cars.
In other words, the auto insurance industry as it exists today is going to be decimated. Self-driving cars, once they gain purchase, should eliminate up to 90 percent of the risk of driving today, which is how many accidents are attributable to human error.
Interestingly, however (and perhaps frustratingly), it may be more expensive to insure self-driving cars in the short-term. That's because actuaries, who rely on large bodies of data to assess risk, which determines insurance premiums, simply don't have access to much data on self-driving vehicles. Hedging their bets, auto insurers are likely to quote inflated rates.
And then, in short order, the whole sector will turn on its head.
Here's a fun thought exercise: What happens to gas stations when a preponderance of cars on the street drive themselves?
There are a few ways to tackle the question. Fleets belonging to ride-share services may not visit commercial gas stations at all if the economics support in-house fill-ups. For privately owned vehicles, it's likely gas stations are going to get a lot more technologically advanced.
In fact, it's a sure bet the pump and your future will communicate with one another and that fill-ups will be executed autonomously, without ever having to press a button or lift a handle. And that could also have a major impact on a mainstay of gas stations, the convenience store. If your car can fill itself up without you, who's going to buy that Snicker's bar?
There's little question that interstate shipping on highways will soon go autonomous. Driving development are major OEMs, which are partnering with technology companies at warp speed to bring autonomous trucks to the road.
For example, Nvidia will provide its autonomous driving platform to vehicle manufacturer Volvo for the development of self-driving trucks. Volvo will use Nvidia's Drive solution to create trucks targeted at freight transport, among other industries.
Behemoth Waymo is pushing into autonomous trucking with testbeds in California and Arizona. The trucking arm of Waymo uses much of the same technology used in its cars and minivans, albeit fine tuned for the challenges of operating a big rig on a roadway.
Even the USPS, historically not the quickest adopter of new technology, has been testing self-driving freight, starting with a route between distribution centers in Phoenix, Arizona, and Dallas.
The shakeup in the trucking industry will have major impacts across the board. Benefits will likely include fewer accidents, more efficient routes, and faster shipping. But employees in the industry may lose out. In fact, the trucking industry may be primed for a major labor crisis. According to a new study, autonomous trucks could eliminate as many as 294,000 driving positions over the next 25 years, including the highest-paying jobs held by senior operators.
State of the technology & regulation
The FAA has been taking methodical steps on the path toward commercial drone usage. Some in the industry groan that it's been dragging its feet.
But slow progress is coming to the US drone sector, including the first approved commercial drone flights beyond line of sight and approval for limited drone delivery in some test beds.
One of those testbeds is suburban North Carolina, where waivers have been granted as part of the FAA's UAS Integration Pilot Program (IPP), which is designed to bring state, local, and tribal governments together with drone operators and manufacturers to accelerate safe commercial drone integration. The program is being seen within the sector as something of a springboard for drone manufacturers.
Israeli company Flytrex, for example, has teamed up with drone services company Causey Aviation Unmanned, Inc. to begin food deliveries by drone in Holly Springs, North Carolina. Flytrex will soon begin on-demand, unmanned food delivery via commercial drones for local residents.
Small steps like this may seem achingly slow, and indeed this is a case where the technology has far outpaced the regulatory environment. But the FAA knows it needs to get drone regulation right lest it open a Pandora's box of unintended -- and perhaps unsafe -- consequences.
That's not to say commercial drones aren't playing an important role in the enterprise already. That's particularly true when it comes to industries that operate largely in unpopulated environments, such as infrastructure or agriculture, or where controlled spaces enable the collection of consent forms, such as golf courses.
And unmanned drones aren't the only kind of self-driving aerial vehicle regulators have been dealing with. It may seem a foregone conclusion that self-driving cars are on the way, but we've heard less about autonomous aircraft.
That's going to change. Following the recent crashes related to failures in autonomous systems on-board Boeing's 737MAX, you might expect consumer confidence to have eroded significantly. However, an ANSYS study found that wasn't the case. In fact, 70% of consumers say they are ready to fly in autonomous aircraft in their lifetime.
More recently, several companies have debuted air taxis, which promise to whisk passengers above traffic en route to their destination.
Infrastructure & agriculture
Drone delivery is still a ways off in the US thanks to the current regulatory environment, but there is an area where drones are already having a massive impact: Inspection.
UAV, which are perhaps more appropriately conceptualized as autonomous flying sensor packages, are perfectly suited for inspecting all kinds of things, from offshore oil platforms to the water needs of crops. Did you know trees are the biggest individual risk to utility lines, poles, and transformers? Trees falling on power lines are the No. 1 cause of power outages and can cause disastrous fires, which is why California utilities have been preempting damage during windy conditions with sweep blackouts.
Historically, the inspection process has included either helicopters carrying LiDAR to captures 3D images of the trees or manual inspection by ground crews. Drones are faster, cheaper, and safer, an enterprise trifecta.
Drones are also the backbone of a host of new techniques known collectively as 'precision agriculture' or 'agricultural intelligence', which is giving farmers new tools to increase productivity while decreasing loss and waste. One company, Taranis, just unveiled a novel weed identification system that combines computer vision, drone imagery, and AI to tell farmers what kinds of weeds are attacking their crops in real time, empowering the farmers to fight back.
In the insurance sector, damage assessments are now frequently performed by drones equipped with machine vision and AI.
It makes sense. When damages occur to homes in the wake of weather events or fire, armies of adjustors are dispatched to physically inspect buildings. Roofs are one of the most commonly damaged components of homes and buildings, and getting up there takes time. Drones make quick work of it, however.
In 2018, The Travelers Companies, a major national insurer, made an investment in drone operations software company Kittyhawk, strengthening the partnership between those companies. Travelers launched its drone operations program in 2016 in the aftermath of Hurricane Matthew.
Logistics & restocking
Drones aren't just for outdoors use; in fact, they're coming to warehouses and fulfillment centers, as well as grocery stores -- from ferrying packages and products from one end of a logistics warehouse to another to general materials handling and shelf scanning functions. Currently useless airspace within factories and stores could become productive thoroughfares. Drones are fast, and they could be added to existing operations without expensive retrofits, which makes them very attractive.
Inventory management is a huge problem in brick-and-mortar stores that boast thousands of SKUs. A company called Pensa is using aerial drones to move its visual perception platform around store aisles. One advantage of drones is that they're far less expensive to customize and deploy than it would be to develop an autonomous mobile robot from scratch.
Another advantage with drones is speed. UAVs can whip around a store much faster than ground robots.
Law enforcement & emergency response
Law enforcement has been keen to adopt drones, which can help monitor crowds, track criminals, and collect evidence far faster than human counterparts.
However, the development has not come without controversy. Nationwide, there's no standard when it comes to police using drones, although the Supreme Court has weighed in generally on the subject in Florida v. Riley. In that case, the Court ruled there's no right to privacy when it comes to police observation in public airspace, reinforcing existing standards when it comes to the use of police helicopters.
In North Dakota, a 2015 law not only makes it legal for police to use aerial drones for surveillance, but also to equip drones with non-deadly weapons such as tasers. Other states and municipalities have been more cautious equipping police with drones. In 2013, Seattle famously reversed course and withdrew plans to equip local police with drones amid escalating protests. The California senate passed a bill that would have required police to secure a warrant for drone surveillance in 2014, but that bill was vetoed by then-Governor Jerry Brown, leaving the door open for police to use drones without warrants.
When it comes to emergency response, however, drones have received broad public support. Drones are playing an increasing role in emergency response, from disaster relief and search & rescue to delivering life-saving medication in remote areas. Recently concluded trials in Ontario have found that equipping drones with defibrillators could save lives by cutting the time it takes paramedics to arrive on scene.
UAV have also been used to locate missing or trapped persons following disasters and to assess damage across wide swaths of land that may be inaccessible to responders on the ground.
Drones have been a boon to filmmakers, significantly lowering costs for aerial shots, which previously required chartering manned helicopters or airplanes. But the ease of access also comes with a downside. Lots of filmmakers are using drones in their work, and oftentimes the shots look cookie cutter, particularly if they were captured with an autonomous drone.
Researchers at Carnegie Mellon University are developing a system for aerial cinematography that learns from human visual preferences in order to enable drones to make artsy filmmaking choices while autonomously filming scenes. The system does not require GPS tags to localize targets or prior maps of an environment.
Drones are also set to become a boon to live sports coverage, although current regulation has put a chokehold on this use case. Fox Sports used an unmanned aerial vehicle during coverage of the 2015 US Open, the first time a drone has been used to cover a major sporting event by a live broadcaster.
To fly the drone near a golf course packed with some 40,000 spectators each day, the network had to secure special dispensation from the FAA. Personnel from FAA were on hand, along with city spotters and law enforcement officers, to keep an eye on the drone's flight path and altitude. In the event of any of several dozen unsafe potentialities -- a boat coming within 1,000 feet of shore, say -- the drone was to land immediately. Permission was only granted because Chambers Bay Golf Course abuts a large body of water, which allowed the drone to safely cover the 16th and 17th holes without flying over spectators.
State of the technology & regulation
Some drones are made to get wet. Autonomous ships and smaller ocean- and lake-going drones are still in their infancy but primed to play important rules in shipping, environmental science, and defense.
If the regulatory environment is a mess when it comes to UAV, it's virtually nonexistent for unmanned aquatic vehicles. That's made the high seas a fascinating sandbox for developers to tinker in.
Logistics & shipping
In May of 2019, a ship called Maxlimer delivered oysters and beer from Britain to Belgium. That may not seem groundbreaking, until you consider that no one was on board during the journey, which marks a first for commercial shipping.
The ship is the brainchild of a company called SEA-KIT, winner this year of the $7 million Shell Ocean Discovery X-Prize. The idea is to develop custom, fuel-efficient ships for infrastructure inspection and potentially cargo applications. Maxlimer navigates busy shipping channels via a remote human operator, but once it's in open water the autonomous equipment takes over.
SEA-KIT is eyeing a transatlantic passage as early as 2020, a feat that could take as long as a month and help prove that international shipping, the engine of global trade, could soon be done largely autonomously.
Aquatic drones are a boon for environmental monitoring. A company called Saildrone, for example, operates a fleet of wind-powered drones that collect data that can be utilized by researchers, including tracking the chemical composition of the water and identifying pollutants.
A consortium of scientists that includes iRobot, called Robots in Service of the Environment (RISE), is developing an underwater robot to kill invasive lionfish by delivering a lethal electric shock.
Another promising development in aquatic drones, which could be massively useful for environmental monitoring, is swarm robotics. A company called, Aquabotix, which makes unmanned underwater vehicles, is commercializing swarm drones for the sea.
The company's latest offering is a small vessel that operates on the surface or underwater. Paired with additional units, the robots can be controlled in a swarm by a single operator as if they were an individual entity, enabling them to capture useful data across wide swaths of water efficiently, quickly, and inexpensively.
Volkswagen's driverless car 'startup' aims to take on Silicon Valley
UPS partners with CVS Pharmacy to develop drone delivery use cases
KT and Hyundai test 5G traffic navigation on autonomous cars
Drones that weave through workers in factories
This autonomous ship aims to steer itself across the Atlantic ocean