If it becomes “fully automatic driving” –.
“Completely automated driving” is almost always the topic of discussion when it comes to dangerous driving by elderly people and the issue of surrendering their licenses. However, when it comes to the prospect of its practical application, it must be said that it is still in the future in all respects, such as technology and legislation.
If we divide the level up to fully automated driving by AI into stages and compare it with the current law, driving assistance for drivers by AI is the current optimal solution. How will AI approach the pressing issue of traffic safety?
Even if cars evolve, the number of accidents caused by human error will not decrease
Dangerous driving by elderly drivers and accidents caused by the sudden illness of drivers have recently become a hot topic in the news, and public interest in traffic accidents is increasing. According to statistics released by the National Police Agency, many of the causes of traffic accidents are caused by human error, such as non-confirmation of safety, inattentive driving, and non-observance of other people’s movements. Although the performance of the automobile itself is improving, there is no end to accidents caused by human error.
In recent years, not only companies owning commercial vehicles but also the general public have become more aware of traffic accidents, and the drive recorder market is expanding rapidly. However, drive recorders are tools for recording driving conditions and checking them later. In the unlikely event that a traffic accident occurs, it will be a “witness”, but it does not directly prevent the traffic accident itself.
The accident rate was reduced by 50-70%.
Under these circumstances, Silicon Valley startup “Nauto” is challenging the “next possibility” of drive recorders. The company has developed a next-generation drive recorder that uses AI to automatically detect dangers while driving in real-time. Mr. Ida talks about its features, saying, “Nauto is a driver safety system that aims for zero accidents.”
Ida: Nauto was launched in the United States in 2017. So far, more than 250 companies in Japan and the United States have introduced it, and as of the end of 2018, the total mileage of the introduced vehicles is over 400 million km.
The major difference between Nauto and conventional drive recorders is that AI detects dangerous driving in real time and encourages drivers to drive safely. An in-vehicle device equipped with cameras for inside and outside the vehicle detects and analyzes the driver’s movements. If the AI detects inattentive driving for a certain number of seconds or more, drowsy driving, or situations where the following distance is not maintained, it will emit a warning sound to notify the driver and notify the operation manager by e-mail.
Using a dedicated web app, the operation manager can check the driving history of each driver and the driving video when dangerous driving is detected, as well as the number of sudden braking, sudden acceleration, and inattentive driving.
In addition, Mr. Ida talked about Nauto’s “value to companies” as follows.
Ida: Many conventional drive recorders record driving images on an SD card, and it was a hassle to insert and remove the card. In addition, many companies had operation managers review all the collected data and check for dangerous driving, which was a heavy burden.
In that respect, Nauto does not require an SD card replacement. In addition, since facial recognition is used to identify drivers, even if one driver drives multiple vehicles, it is possible to collect operational data in real-time, such as which vehicle the driver drove for how long. In addition, only videos of dangerous driving are uploaded to the cloud via mobile communication, so the operation manager does not have to check all the operation videos.
Digitization by AI can significantly reduce the time and effort of vehicle management for companies. Furthermore, since the driving situation of each driver is visualized, it can be used for more efficient safe driving guidance.
Our main customers are transportation companies, taxi companies, and companies with commercial vehicles. Above all, it is a company that has introduced a drive recorder, but has experienced that “operation was difficult” and “accidents did not decrease”. Companies that have introduced the system have not only reduced the time and effort involved in vehicle management but have also reduced the accident rate by 50-70% in some cases. Data has shown that inattentive driving by drivers can be reduced by more than 50% after about six months of operation.
While some drivers say they are happy with the increased safety, some people are initially reluctant to have a camera in the car. However, in most cases, if you explain that data is not sent to the operation manager if you are driving safely and that the design is designed with privacy in mind, they will be convinced.
Advanced technology from Silicon Valley. AI accuracy is already on par with human eyes
So, what level of accuracy has AI reached in judging dangerous driving? With Nauto, AI can detect distracted driving and tailgating with as much accuracy as the human eye can see.
Ida: In conventional solutions, the mainstream was pattern recognition, in which an alert was sounded when the eyes deviated from the reference point by several centimeters. In this case, adjusting the angle was difficult and accurate data could not be collected.
Nauto, on the other hand, uses algorithms that use deep learning. AI automatically detects where the person is and determines how the person is moving. Since the accuracy is similar to that seen by humans, it is possible to detect dangerous driving without problems even when wearing sunglasses or in dark conditions such as at night. Customers have also praised it for its high accuracy.
The system is developed in the home country of America. However, the road environment and human behavior patterns differ between the United States and Japan. It is said that Nauto is localizing according to the traffic rules and road widths of each country.
“Human-focused data” is necessary to realize fully automated driving
At Nauto, AI learns from the vast amount of operational data it collects, creates new algorithms, and updates its software once every 90 to 120 days. It is unique to the cloud service that it is not the end of the product once it is released, but it continues to be updated. However, beyond the framework of simply updating products, Nauto is working toward the utilization of fully automated driving in the future.
Ida: Various services will appear in two to three years for autonomous driving under limited conditions, such as within a specific site, but the realization of fully autonomous driving is still a long way off. There are many hurdles to overcome.
For fully automated driving, AI must be able to know the driver’s situation and predict what the driver will think and how he will act. For that purpose, we need to collect a huge amount of data and perform simulations. Collecting such “human-focused data” through AI drive recorders is an important role for us, and we believe it will be a big step toward the realization of fully automated driving.
Naruto’s mission is to use AI and image recognition to realize safe transportation. It is said that the development and sales of AI Drive Recorder are just the starting point.
Ida: There is a big gap between the current situation surrounding drivers and fully automated driving. Our job is to fill in the gaps. While collecting data and conducting simulations for the realization of fully automated driving in the future, we will provide drivers with a safe and efficient driving environment in a society where humans and automated driving will coexist for some time to come. This is the core of our business.
Furthermore, in the future, we are also considering the development of various data businesses using the data collected by AI Drive Recorder. We believe that there are various business possibilities if we take advantage of the feature of collecting information outside the vehicle just by driving and having AI analyze it.