The Path to Regional Revitalization Through Autonomous Driving

Japan currently faces an ultra-aging population and declining population, and maintaining the gross national product is an urgent task. In today's society, a large amount of energy and time are consumed for the movement of people and the transportation of goods, so the development of transportation technology is indispensable for maintaining the gross national product. Autonomous driving is expected to serve as an underlying technology. In collaboration with domestic automobile manufacturers, telecommunications carriers, government ministries and agencies, etc., I have been conducting research and development on autonomous driving for automobiles from 2016 to the present time. In this article, I will explain the contents of the Best Paper Award which I won at the international conference IEEE IoTNAT 2021 for joint research results^[1] with NTT and the Ministry of Internal Affairs and Communications.

I was born and raised in a small rural town near the mountains in the northern part of Yamaguchi Prefecture. My father was a driver for a route bus. When I was a child, I enjoyed talking with my friends from elementary school while going to school in the bus driven by my father. Also, even as a child, I remember feeling proud of how my father safely delivered the many people riding his bus to their destinations. Since then, I have witnessed the current status of public transportation in rural areas by observing my father's work as a route bus driver.

The population in rural areas is declining rapidly, and there has been a drastic decrease in the number of public transportation users. In 2022, local bus companies have become unable to maintain their operations without subsidies from prefectures and municipalities. Basically, none of the bus routes are profitable, so buses come at an interval of one or two hours. This is very inconvenient. Additionally, the fares of route buses are increasing every year in conjunction with rising gasoline prices. Furthermore, it is impossible to increase the salary of bus drivers amidst these worsening business conditions. As a result, a sufficient number of bus drivers cannot be recruited and there is a chronic labor shortage of route bus drivers.

In rural areas where public transportation has deteriorated due to the declining birthrate and aging population, ownership of a private vehicle is essential for daily transportation. This is a factor that increases the number of elderly people who cannot voluntarily surrender their driver's license even if they no longer wish to drive. If we are unable to take measures to address the current state of public transportation in rural areas, the number of elderly people who want to quit driving but cannot will continue to increase. The increase in traffic accidents caused by automobiles driven by the elderly is one of the social problems facing Japan.

In today's world of active exchange of people and goods, people will continue to leave rural areas where public transportation has deteriorated, and will flow into urban areas with more convenient transportation. If we want to achieve regional revitalization, we must eliminate the inconvenience of traveling in rural areas. This is the reason why I began to research autonomous driving.

I would now like to start a technical explanation related to autonomous driving. However, similar to rocket science, autonomous driving is a collection of advanced underlying technologies from various engineering fields, and it is difficult to explain all of those fields from a bird's-eye view. Therefore, based on my background in information networks, I focus on two elemental technologies which I consider to be particularly effective for the realization of autonomous driving.

The first underlying technology is connected cars^[2]. A connected car has the ability to be constantly connected to the Internet. Connected cars make it possible to collect various data such as the condition of automobile components and surrounding road conditions, and then to analyze those data on an external computer via the Internet. Some readers may think that since a car is equipped with a battery, it would be more efficient to perform calculations using a computer installed inside the car. However, in reality, this is no easy task. Autonomous driving requires an enormous amount of computer resources, while automobile batteries are weak. Once, I naively asked an engineer of an automobile manufacturer to use a car battery to power a computer that runs artificial intelligence for autonomous driving. With a bitter laugh, the engineer replied that the battery would be drained in one second. Today, this is an important memory for me. Simply installing a computer in an automobile is not sufficient to achieve autonomous driving with a high level of safety. Instead, the best method is to use the Internet to achieve a connected status for computer resources such as the cloud and the autonomous driving car.

The second underlying technology is a cyber physical system^[3]. A cyber physical system builds a virtual world (cyber space) in the cloud by collecting and analyzing various data in the real world (physical space). In the case of autonomous driving, information such as connected cars, pedestrians, traffic lights, buildings, and weather is collected from an actual city area, and a virtual city area simulating the real world is constructed in the cloud. By utilizing the cyber physical system for autonomous driving, it is possible to respond to pedestrians suddenly emerging from blind spots that cannot be recognized by a single connected car. Furthermore, multiple connected cars can share information and engage in mutual cooperation in order to further optimize the entire urban transportation network. Cyber physical systems are required to exchange data in real time while maintaining high communication quality in status where the connected car and cloud are constantly connected. Therefore, in recent years, the role of networking technology for connecting connected cars and the cloud is being treated with greater importance than ever. For example, there is increasing cooperation between automobile manufacturers and telecommunications carriers.

In the cyber physical system, by performing a detailed simulation for the state of a connected car driving in an urban area, it is possible to achieve safe autonomous driving through cooperation among multiple connected cars. To achieve this, CAN data^[4] must be extracted from inside of the connected car and delivered to the cloud within an extremely short time of 10 to 100 ms. CAN data is important data that represents the state of the engine, steering wheel, brakes, etc. of the connected car, and it is indispensable data for simulating the running state of the connected car.

On the other hand, real-time collection of CAN data is known to be one of the most difficult technical issues in research related to autonomous driving. The reason is that the size of CAN data is extremely small at 16 bytes, and more than 1,000 instances of data are generated per second per connected car. An enormous number of small CAN data from many connected cars running in the city will flood the cloud. However, current networking technology cannot collect large amounts of CAN data in the cloud. To explain by using a familiar example, it would be an inefficient situation in which "1,000 erasers ordered by mail are individually wrapped in cardboard and delivered 1,000 times by different delivery companies."

In response, by using a mechanism called edge computing^[5] which is promoted by telecommunications carriers in 5G, I invented a technology for real-time compression of CAN data transmitted by multiple connected cars. Using the technology that I created achieves an efficient delivery situation in which 1,000 erasers ordered by mail are packaged together in a single cardboard box and batch delivery is performed by a single delivery company. When I actually drove a connected car and verified the effectiveness of the developed technology, I verified that it could achieve a compression ratio of 88%.

By using this technology, it will become possible to realize a cyber physical system for autonomous driving at low cost, even in rural areas. Moving forward, if we can make it easier to introduce autonomous driving, it will be an effective method for solving the problem of movement in rural areas. Although there are still an enormous amount of issues to be solved for autonomous driving, the technology of autonomous driving is steadily evolving every day. I hope that one day, even in my hometown, the sight of self-driving buses will become commonplace. At that time, I would like to ride together with my father on a self-driving bus and take a journey to somewhere far away. Along the way, I'm sure that I would have to listen to my father brag that he drives better than a self-driving bus, but I would be happy to listen to him talk.

(Related Literature)

[1] M. Yoshida, K. Mori, T. Inoue, H. Tanaka, "EdgeRE: An Edge Computing-enhanced Network Redundancy Elimination Service for Connected Cars," in IEEE IoTNAT 2021, 2021 (Best Paper Award).
[2] Ministry of Internal Affairs and Communications: WHITE PAPER Information and Communications in Japan 2015, "Chapter 4: ICT and Future Lifestyles-Connected vehicles and autonomous vehicles" (referred to on May 12, 2022).
https://www.soumu.go.jp/johotsusintokei/whitepaper/ja/h27/html/nc241210.html
[3] Ministry of Internal Affairs and Communications: WHITE PAPER Information and Communications in Japan 2020, "Chapter 4: Beyond 5G-Cyber Physical Systems" (referred to on May 12, 2022).
https://www.soumu.go.jp/johotsusintokei/whitepaper/ja/r02/html/nd140000.html
[4] International Organization for Standardization, "ISO 11898-1:2015 - road vehicles -- controller area network (CAN) -- Part 1: data link layer and physical signalling," https://www.iso.org/standard/63648.html, 2015 (referred to on May 12, 2022).
[5] European Telecommunications Standards Institute: ETSI ISG MEC 2022 (referred to on May 12, 2022).
http://www.etsi.org/technologies-clusters/technologies/multi-access-edge-computing