What it is
Just like in airplanes, Autopilot is a driver assistance system, designed to make (especially long-distance) travelling less stressful. However, that means, that like in airplanes, the driver is still responsible and should be able to take control at any time. The 1st generation system is very good for motorway driving and traffic jams.
What it includes
Autopilot is basically a bundle, consisting of several sub-systems / features, which serve different purposes.
#1 Traffic Aware Cruise Control: Can be used alone or in combination with Autosteer. Works like a regular cruise control, but can automatically lower the speed to match the speed (and keep safe distance, which can be manually adjusted) to the preceding vehicle.
#2 Autosteer: Automatically keeps the car inside of a current lane. Recommended to use only on roads with clear lane markings, no steep turns, and a central divider. (Auto lane change is an extension, which allows the system to automatically change lanes when driver activates a turn signal.)
#3 Autopark: Continuously scans the area around the car at lower speeds to search for an adequate parallel / perpendicular parking space. Once a space is found, an icon appears in the instrument cluster. Once the driver confirms, the car can park into a given space without any other input. (A parking space is defined by two vehicles / objects with a free space between them.)
#4 Summon: Works with the driver outside of a car. Key fob / a mobile smartphone app can be used to move the car forward / backward to for example get out of a tight parking spot.
How it works
The system works best on roads with clear lane markings, no too sharp curves, and a central divider. This is because Autopilot uses different components to determine what’s happening around the car.
#1 Camera in the front windshield: Looks for lane markings and vehicles in front of the car. It can also recognize speed limit signs.
#2: Radar in the front bumper: Does pretty much the same thing as the camera, but it can better estimate distance and speed of given vehicles (even in less than ideal weather), so the data gets combined to create a more complete image for the on-board electronics to use.
#3 Ultrasonic sensors around perimeter of the car: Sense close objects, like barriers right next to the car, and other vehicles in your blind-spots. These sensors are also used for finding parking spaces.
#4 GPS and high resolution maps: Used more as a back-up when sensors mentioned above do not get 100 % sufficient information.
All of these sensor feed the data into a computer, which then decides how to act. It is able to control the braking, acceleration, steering, and “gear” selection.
What it’s not
This definitely isn’t a fully autonomous car, which can drive itself without the driver paying attention to the overall situation on the road. Even though it may be tempting to do something else behind the wheel when Autopilot is active, one never knows if an unexpected situation occurs. In that case, the driver needs to be aware of the situation and react immediately, which is not possible if he doesn’t pay attention to the road.
Solution for the general consumer
The fact why Autopilot may seem so controversial is because it is labeled as “BETA”, which is very unusual in the automotive industry and marks something not totally finished / bug-free in the computer world. However in this case, it should suggest that a human driver should not fully trust the computer and activates these features knowing, they may not work in all possible scenarios.
To further make this clear, I would suggest Tesla to create a short instructional video about the functioning and limitations of Autopilot. This would need to be watched prior to enabling the Autopilot features. After that, the person watching the video should take a short test / quiz and only be allowed to take advantage of Autopilot if all of the given questions were answered correctly.
This way, people could to better judge when to use the help of this assistance system.
As of late 2016, every new Tesla rolling off the production line gets a new set of hardware as standard. The 2nd generation Autopilot uses: 8 cameras, a front facing radar, 12 (more refined) ultrasonic sensors, and more powerful computers for data processing.
Elon Musk has stated, that it will be possible to achieve Level 5 autonomy with this hardware just through software updates, meaning you could call the car to come and pick you up from across the city and drive you back home, even after a couple of glasses of wine.
And the Ticking Time Bomb
In May 2016, Autopilot may have been active during a crash, which was fatal. This is a first such situation. Two opposing views emerged.
#1 Pessimistic view: Tesla shouldn’t offer the Autopilot as a “BETA” feature, as it is not completely polished. The users are put at risk and are Guinea pigs for Tesla.
#2 Technology lover’s view: Autopilot has had its first fatal crash in 130 thousand miles, making it safer than the US average (1 death per 100 thousand miles), and worldwide average (roughly 1 death per 60 thousand miles).
The truth is somewhere in between. The fact is that Tesla doesn’t use their customers as Guinea pigs, as it is up to them to enable Autopilot features, which are off by default. And when it comes to statistics, Autopilot hasn’t driven enough miles yet, to make an accurate statement. The thing which could prevent further development of autonomous cars is thrust. Or better said, the lack of it. If media pipe out headlines, that a human has been killed by a car, it can make these cars look dangerous. But that’s like saying “flying is dangerous, because there has just been an accident”.
I think I have written enough. Now, it’s up to you to make up your mind and form view #3.