A self-driving "panda" test vehicle built for Google's Self-Driving Car Project in 2014

The Google Self-Driving Car Project's "panda" test vehicles became its trademark. Now named Waymo, the project has retired the panda cars in favor of sturdier on-road test vehicles. Image source: Alphabet.

There has been a lot of talk -- among investors, among futurists, and definitely among people in the business of building cars -- about autonomous vehicles.

It's very likely that the first fully automated vehicles will come to market in just a few years, and maybe earlier. Here's what you need to know now to understand what's coming and when.

A Chrysler Pacifica minivan with self-driving sensors and Waymo logos on a public road in California

Waymo is now testing its self-driving system with a fleet of specially modified Chrysler Pacifica Hybrid minivans. Image source: Waymo LLC.

What are autonomous vehicles?

An autonomous vehicle is one that automates the task of driving -- in other words, it doesn't require a human driver to get from place to place. It could be anything from an electric scooter to a tractor-trailer truck. Autonomous vehicles are sometimes referred to as "driverless" or "self-driving" vehicles -- all three terms have the same meaning.

An automated convoy of Mercedes-Benz tractor-trailers in a demonstration on a German highway last year

Self-driving technology has applications beyond passenger cars. German auto giant Daimler is one of several companies working on self-driving heavy trucks. Image source: Daimler AG.

How do autonomous vehicles work?

The self-driving systems currently under development use a combination of sensors, including cameras, radar units, GPS, and lidar (a radar-like technology that uses invisible lasers) to assess their surroundings from moment to moment. That information is fed to a sophisticated computer "brain" that knows exactly where the vehicle is located (often down to a few inches) and figures out how to drive in traffic and navigate safely around obstacles.

Many of the systems under development use highly detailed 3D maps to help the computer navigate, or as a safety check to ensure that the computer really knows exactly where the vehicle is located. The limitation: Vehicles with those systems often need a human driver to take over when they go beyond the edges of their maps.

(Here's an article where you can learn more about the different levels of vehicle automation, from driver-assist systems to full-blown self-driving.)

Ford Fusions with self-driving hardware traveling on a snowy road

Ice and snow present special challenges to a self-driving car's sensors. Ford is one of several companies that has stepped up its testing in winter conditions. Image source: Ford Motor Company.

Self-driving systems face big challenges

Self-driving systems consist of hardware (like sensors) and software. For the most part, the hardware is ready: Tiny cameras, radar units, GPS modules, and computer processors are all off-the-shelf technology. Lidar systems also exist, but need further refinement before they're cost-effective for mass-production. (That's happening now.)

Getting the software to the point where we know that it's safe enough for widespread use is the challenge. Extensive testing will be required. That's why the automakers and software companies developing self-driving systems are deploying test fleets on public roads: to rack up test miles, especially in situations where driving can be challenging (in big-city traffic, for instance, or in snow and ice).

These test vehicles have human engineers on board to grab the wheel if needed and to make observations on the vehicle's behavior that will help refine the systems. Most experts (inside and outside the industry) think that it'll be a couple more years before the first systems are ready to hit the roads. Of course, it's hard to say when government regulators will agree.

A Volvo SUV with Uber logos and self-driving sensors on a street in San Francisco

Ride-hailing giant Uber Technologies is working on its own proprietary self-driving technology. Its fleet includes Volvo SUVs and Ford Fusion sedans. Image source: Uber Technologies.

Autonomous vehicles have raised ethical concerns

Self-driving developers are constantly considering "what if" scenarios that raise questions about the choices a self-driving vehicle's computer should make in different kinds of emergencies. Humans make these kinds of split-second decisions all the time, but it's hard to give a computer the nuances of human experience that will lead it to make the right choices -- even when we can agree on what "the right choices" are.

It's likely that the answers to many common scenarios will be standardized (and maybe governed by regulations) in time. But getting there is one of the challenges facing self-driving systems.

Chevrolet Bolts with self-driving hardware move down an assembly line at a GM factory in Orion Township, Michigan.

In a first for the auto industry, GM recently built over 100 self-driving Chevrolet Bolt test vehicles in one of its regular factories. Image source: General Motors.

The first autonomous vehicles are a few years away

Automakers like General Motors (GM -2.72%), software giants like Alphabet's (GOOGL -0.02%) (GOOG 0.10%) Waymo self-driving subsidiary, companies that supply technology to automakers like Delphi Automotive (DLPH) and NVIDIA (NVDA -2.61%), and other entrants like Uber Technologies are all either putting their own self-driving test vehicles on the road now, or supplying key components to others that are.

Most think that the self-driving systems that rely on those 3D maps are just a few years away from coming to market. Fully self-driving systems that can go anywhere a human driver can will likely take longer.

Mary Barra is speaking while standing next to a Chevrolet Bolt EV with self-driving sensors.

GM CEO Mary Barra has hinted that her company's first self-driving vehicles will be for fleet use. Lyft is a likely first customer; GM owns 9% of the ride-hailing start-up. Image source: General Motors.

You may not be able to buy a self-driving car at first

While a few companies -- like Tesla (TSLA -4.02%) -- are planning to bring self-driving technology to retail customers, others expect to offer it to fleet customers (like Uber or rival Lyft) first. That's a way to deal with the limitations of the maps: Lyft's self-driving vehicles might be programmed to stay within a city's limits, for instance. It might be a few years before those companies offer the technology directly to consumers.

But it'll happen eventually. At least as of right now, it seems likely that self-driving functionality will roll out gradually across the auto industry over the next decade or so.

An Audi SUV with self-driving hardware and Delphi logos

Auto supplier Delphi Automotive is working with Intel on a self-driving system it plans to offer to automakers late in 2019. Image source: Delphi Automotive.

The technology will be a commodity before long

The tech media loves to talk about the "race" to develop self-driving technology. But there are a lot of different efforts underway, and several are thought to be close. The most likely scenario is that the map-limited self-driving technology will be available to any automaker that wants it within three years -- at a cost.

At that point, it'll be a commodity -- though it might be several more years before prices come down to the point where it's widely available.