US experts: True driverless cars won't be available until 2075

US experts: True driverless cars won't be available until 2075

Autonomous driving systems are already being installed in modern cars, but they can mostly only assist in driving under certain conditions. Even though automakers and consumers are eagerly awaiting fully autonomous cars, there are still many different stages of automation to go through in order to truly achieve this goal.

Driving a vehicle is far more complex than people think, and it involves many skills and behaviors, some of which are very easy to automate, such as using the experience accumulated in traditional cruise control systems over the past few decades to allow cars to automatically maintain a constant speed on the highway.

With the advancement of technology, engineers have also successfully automated some other tasks: the currently widely used adaptive cruise control system can maintain appropriate vehicle speed and distance between vehicles; the lane keeping system is also used in new models of Mercedes-Benz and Infiniti, which can coordinate the information collected by cameras and sensors and use the steering control to keep the vehicle in the middle of the lane. But even so, there is still a huge gap to cross from the current car automatic system to fully automated driving.

The Society of Automotive Engineers International defines five stages of autonomous driving. The first three stages of technology all rely on humans to handle emergencies during driving, and the third stage system allows drivers to switch to autonomous driving in specific scenarios, such as when stuck in traffic on the highway. The fourth stage system can handle all driving-related tasks, but the use scenarios are strictly limited to closed parking lots or dedicated lanes on highways. The top fifth stage is a fully automated car.

Regardless of what the public thinks, human drivers are very good at avoiding serious crashes. According to 2011 U.S. traffic safety statistics, for all drivers, there is approximately one fatal collision every 3.3 million hours of driving and one injury collision every 64,000 hours of driving. These numbers set an important safety goal for autonomous driving systems, which is that the minimum safety standard for autonomous driving must not be lower than that for human driving. There is still a long way to go before autonomous driving can achieve this reliability, and the distance is longer than many autonomous driving enthusiasts think.

Think about how often your laptop crashes. If such software is used for car driving, the "blue screen of death" will not only be a complaint, but a one-tenth-second delay in software response may cause a traffic accident. Therefore, the software involved in autonomous driving must be designed and developed using completely different standards and cannot be based on existing devices on the consumer market. It is extremely difficult to meet these standards, and fundamental breakthroughs in both software engineering and signal processing are required.

When driving, there are many vehicles in front and behind, and obstacles may appear suddenly. In any case, the car's autopilot system needs to make decisions within a few microseconds. Therefore, compared with the autopilot system and code used on airplanes, this system is several orders of magnitude more complex.

I would say that it is difficult to achieve a fully automated driving system by 2075. Will it be achieved earlier than this time? It is possible, but not much earlier.

We can imagine that in the next decade, there will be an automatic parking system that will allow the driver to get out of the car at the entrance of the parking lot and let the car automatically enter the well-equipped parking lot that does not allow pedestrians and non-automated cars to enter. Low-speed driverless passenger cars are also suitable for pedestrian areas, commercial areas, university campuses and other places in cities where high-speed vehicles are not allowed to pass.

In these environments, sensors with limited performance can effectively detect pedestrians and cyclists.

These scenarios may not be as futuristic as having a personal electronic driver, but the technology is definitely available, and it may even be available soon.

(Steven Stoffel is one of the pioneers of intelligent transportation systems in the United States. Special contribution to Global Science)

<<:  2017 Global Innovation Report: Pan-AI fields gain strength, while genomics/VR cool down

>>:  VR glasses startups are booming, and these manufacturers are really good at it

Recommend

Is it really necessary to clean barnacles from whales and turtles?

Recently, there have been many videos of people r...

Changba: Four basic ways to attract users: mining, support, output and retention

What are the users of Changba like? There are sli...

International Volunteer Day丨Do you know these things about volunteer service?

I am Dongdong Miao, International Volunteer Day i...

Cats or dogs, who has a better sense of smell?

Cats or dogs, who has a better sense of smell? Au...

How to quickly and effectively increase the number of APP users?

In recent years, more and more people have entered...

Let you know the development history of speech recognition technology

Author: Yang Jun, Unit: China Mobile Xiong'an...

3 mind maps to help you learn how to plan and operate community programs!

This is a simple mind map of paid community plann...