The Case of the Google Driverless Car

Google was devoted to developing the fully self-driving car, and even emphasized

“driver-less”, because Google believed the partial-automatic driving mode may make the drivers deconcentrated. The research and development of the Google driverless car included two types of vehicles, one of which was an existing vehicle such as a Toyota Lexus embedded with the autonomous driving system (Google, 2016), and another one was the original 2014 prototype of the Google driverless car without a steering wheel, brake pedal and gas pedal and only one button (Retrieved 14 May 2017, from https://www.youtube.com/watch?v=ftouPdU1-Bo). The original driverless car prototype began road tests in 2015 and had accumulated 3,000,000 miles as of May, 2017 (Retrieved 14 May 2017, form https://waymo.com/ontheroad/). To explore the driverless car without human intervention, this research focused on the latter Google driverless car.

Use Case Diagram

From the use case diagram of the Google driverless car shown in Figure 9, we determined that the Google driverless car realized level 5 (full automation) of the SAE J3016™ classification system in which IT possessed high autonomy ability. The main user of the Google driverless car was a passenger rather than a driver. Passengers could summon the car by their smartphone, and it would automatically drive to the passengers’

location. After getting in the car, passengers decided the destination and pressed the start button to enable the automatic driving. However, during the driving process, passengers did not need to pay attention to driving decisions and conditions, and could do something else such as resting and working. In an emergency, passengers could press the button to make the car safely stop in the shortest amount of time. In addition, the Google driverless car also provided useful information such as weather reports (Google self-driving car project, 2014, retrieved 20 April 2017, from https://www.youtube.com/watch?v=CqSDWoAhvLU; TIME Magazine, 2015, retrieved June, 6, from https://www.youtube.com/watch?v=ftouPdU1-Bo; Google self-driving car project, 2015, retrieved 20 April 2017, from https://www.youtube.com/watch?v=uCezICQNgJU; Waymo, 2017, retrieved 20 April 2017, from https://waymo.com/). Among the use cases, three were related to driving decision circumstances: deciding on the destination, starting the autonomous driving

30

and stopping the car in an emergency.

Moreover, three critical data sources were shown on the right-hand side of the use case diagram. The map service provided a map and location information to position the car and use for route planning before driving. The second data resource was detectors that collected numerous surrounding data in real-time for driving decisions. The last one was the road information including the real-time road conditions from the Internet and other vehicles.

Figure 9. The Use Case Diagram of the Google Driverless Car

Framework of the IT role – Autonomer

Based on the above use case diagram, we could apparently observe that people did not make decisions during the driving process except for emergencies; therefore, even blind people could “drive” alone with a Google driverless car.

We applied our research framework to the driving decision process of the Google driverless car and generated the result shown in Figure 10. In the beginning, the Google driverless car had to identify the decision problems such as how to move in the next milliseconds. Next, the Google driverless car collected and analyzed the data from cameras and detectors including the nearby objects/obstacles to generate action alternatives such as adjusting the speed, changing the lane or both. Then, it decided how to drive and control the car. At last, the action performance would be tracked by the Google driverless car to adjust the decision model for a better subsequent decision. The framework of the Google driverless car differed from Automation, Supporter and Mentor. In this framework, IT apparently dominated each step of the decision process.

31

According to the use case diagram, we could observe that users did not need to decide how to drive the car such as by changing the speed and direction. The Google driverless car dominated the entire driving decision process without any human intervention. The Google driverless car collected and analyzed data from the Internet and detectors and made decisions by itself in real-time to address dynamic driving decision circumstances. To handle unpredictable incidents that were not written into programs in advance, the IT of the Google driverless car required not only decision-making ability but also self-learning ability, such as a human would.

From the framework of the Google driverless car, there were two big differences compared to the other IT roles. One was being the “decision maker” such that IT could determine by itself without human intervention; and the other one was being the

“learner” such that there was more feedback than ever before, which meant continuous tracking, adjustment and growth. That is, IT could not only independently build models based on historical data and human preference but also predict the unknown. Because of the learning ability and decision-making ability, the new IT role had a high-level autonomy for judging. This was the difference compared to “automation”; therefore, instead of classifying the IT role as one of automation, we define such an IT role as an

“Autonomer” which means the IT possesses high-autonomy to learn and judge.

Figure 10. The Framework of Autonomer Type IT

(The topmost is the original framework, the lower two marked the ability of IT)

32