Riding in a self-driving car for the first time is usually a vaguely nerve-jangling experience. Despite the autonomous cars’ typically flawless performance in demos and the resulting rush, most first-time riders—especially control-freak drivers—don’t immediately trust the machines with their lives, not out in traffic at speed. Not just yet. It’s a process.

You can expect that notion to persist for the next fifteen years or so because it’s going to take that long for true autonomous vehicles—that is, bona fide robot cars that deliver bulletproof safety and reliability in all conditions—to hit the road. Even that will just start the process of convincing owners and drivers to buy in, initially via autonomous taxis and rentals, that will allow the technology to penetrate first premium, then mid-priced brands, and eventually the rest of the global car fleet. Paradoxically, however, most newer cars already contain many of the technological building blocks that will at some point combine to supply substantial degrees of driverless function for at least part of the time in the not-too-distant future.

“Although we’re closer than ever to bringing self-driving cars to market, today’s technically impressive demos of autonomous vehicles by companies like Google and Mercedes-Benz can be misleading,” said Cosmin Laslau, an analyst at Lux Research, a technology business research and consulting firm who just authored a report on the economic prospects of robotic cars. Those cars, he said, are limited to certain routes and areas that have been pre-scouted and thoroughly mapped out—‘crawled’ in the Web sense—to significantly lighten the data processing burden. Autonomous operation is likewise curtailed by bad weather, visibility, and road conditions.

“Those cars have what NHTSA defines as Level 3 part-time autonomy, not the full Level 4 capability, which is safe self-driving all of the time in all conditions,” he explained. “No one has demonstrated Level 4 autonomy yet.”

Crashless cars

For cars to make the jump to Level 4 autonomy, they need to be essentially crashless. With human error implicated in 90% of 1.3 million road deaths that occur each year worldwide, the prospect that a car maker will one day be able to claim that no drivers will die in a new model is something to think about, Laslau pointed out. Add in the millions who would avoid injury and the whopping $500 billion in additional savings that the World Health Organization estimates could accrue from curbing costly crashes.

Commuters will benefit greatly from the convenience driverless cars offer, of course. The average U.S. commuter spends up to 200 hours a year behind the wheel so the reduction of travel tedium and all the extra time will be big. Self-driving cars could also cut traffic congestion by enabling tighter use of existing road space, but also if companies like Zipcar and Uber can provide novel shared-use programs for employing cars and available parking spots.

Self-driving cars will, in addition, lead to noteworthy efficiency gains. A 10% fuel-economy boost is expected from smoother automated driving, and should cars and trucks make use of aerodynamic drafting in tightly spaced road trains and platoons, their fuel savings could rise another 10% to 30%.

Laslau and his colleagues interviewed participants in the nascent value chain that is coalescing around self-driving cars, including OEMs, hardware and software suppliers, government regulators, and insurance executives to evaluate the commercial readiness of the technology and prospective enterprise. The Lux team then assembled a market model and timetable and forecasted future performance.

Five levels of autonomy

You probably drive at least a Level 1 car, one that contains as much as $500 of basic anti-lock brake and stability control driver-assist systems, whereas your parents drove Level 0 cars. More modern Level 2 cars have advanced driver-assist features such as adaptive cruise control, lane-departure warning, and automatic emergency braking—relatively complex functions that require coordination among sensors as well as some control smarts. These relatively modest self-driving systems, which cost from $1500 to $3000, are steps along the road to some greater degree of autonomy.

Level 3 cars can self-drive for at least three quarters of a journey, Laslau said. During an average commute, for instance, autonomy might take over on well-known local residential and workplace streets, highways, and in stop-and-go traffic, but not in poor weather or areas that are not well mapped. Unfortunately, today’s economic reality is that such capabilities could cost from $25,000 to $150,000.

Self-driving technology will be slow to make a commercial impact, according to the Lux view. Level 2 vehicles are predicted to dominate until 2030, rising from 3% today to 57% in 2020 and 92% by 2030, when only 8% of cars will reach Level 3. “And even then there will be no Level 4 cars,” he said.

The Prius of autonomy

By 2030, the burgeoning value chain for self-driving systems will provide revenue totaling $87 billion, the model says, with software, computer hardware, and key sensor technologies leading the way. The design choices associated with the arrival of the autonomous car—particularly those in proprietary software, the field’s secret sauce—will be key brand differentiators for automakers and their suppliers, Laslau said, warning that those that lag—or pick wrong—will face dire consequences in market share and brand image. Choosing correctly, on the other hand, could result in something resembling the "Toyota Prius of autonomous vehicles."

The market model indicates that the software industry will reap the lion’s share of economic activity in 2030, a $25-billion segment. Algorithms that are largely invisibly to the driver process machine vision and radar-ranging information and fuse the sensor inputs to perceive objects and obstacles through inference and interpolation. This coherent virtual map of the vehicle’s environment can then be navigated by other programs using enhanced GPS and functional AI (artificial intelligence) systems, plotting safe pathways through the world.

On the hardware end of the market, the model says that suppliers of optical sensors will do about $8.7 billion in business while radars should generate $5.9 billion by 2020 because they both appear in current Level 2 cars. Computers are expected to take over the hardware lead in the following decade because of the rising processing demands of Level 3 autonomy, which is a $13-billion opportunity.

Sensors, which can be expensive, need to be low-cost, redundant, and highly reliable. They also have to be well integrated with the software that needs to respond in real-time to ensure safety. Laslau noted that only seven of the 74 vehicles tested for crash-avoidance competence by the Insurance Institute for Highway Safety in 2013 could handle hazards that occur at 40 km/h (25 mph). In addition, external help with the information processing load from V2V or V2I is not at all assured.

Beyond the technical difficulties, the analyst said, numerous open questions face self-driving vehicles, including uncertainties over consumer acceptance, insurance and liability responsibilities, the emerging value chain, as well as a lack of clear, consistent government regulations.