Focus on establishing big data is required, including detailed Prior Map, real-time sensing data, etc.

“Google’s patents deal with software design ability regarding AI, machine learning (ML) and deep learning (DL), which are core factors of autonomous cars, as well as real situations during road tests, coping methods and learning.” Dr. Won-Yong Cha, Korea’s representative futurologist and director of ASPECT, explains thus in his recently-published book ‘Focused analysis of Google’s AI type autonomous vehicle patents.’

Total driving test distance that Google’s autonomous cars have covered to acquire relevant patents is 3.85 million kilometers (2,408,597 miles), among which distance driven in autonomous mode is 2.27 million kilometers. This accounts for around 60% of total distance and Google increased the figure up to 80% last year.

 
▲Google proceeded with road test of its 55 self-autonomous cars in two cities of two states from 2009 to January 31, 2016 (Photo courtesy of Google)

The autonomous mode of Google’s self-driving cars has been deactivated 341 times during the test period, and most of which (304 times) occurred on streets. What caused their deactivation were street incidents such as weather conditions on the roads or jaywalking. “What is important is that Google found out the reasons for deactivation of autonomous mode and had its autonomous cars learn the situations in virtual lab, which resulted in the company’s relevant patent applications,” said Dr. Cha. He pointed out that “if very powerful simulator is available for self-autonomous car, the learning can be done even in garage. Inside virtual lab of such a simulator, Google’s self-autonomous cars are learning while driving 4.8 million kilometers (3 million miles) of virtual road a day.”

Of course, to enable autonomous driving, detailed prior map based on big data and detailed real-time data must match each other. “For this comparative analysis, such big data must be established beforehand. This is the very reason Google did real road test for 3.85 million kilometers,” he explained.

 
▲Autonomous car contest held in Taegu Intelligent Automotive Parts testing ground last year (Ministry of Commerce & Industry).

Urgent need to establish ecosystem through platform

Dr. Cha pointed out that, with Google developing complete form of autonomous vehicle using AI, we need to urgently establish ecosystem through platforms of domestic conglomerates, small and medium companies and venture companies. In the introduction to his books, he states that “Korea needs to invest in autonomous vehicles and road infrastructure from a long-term perspective as USA does. Even though R&D of autonomous vehicle itself is important, we should not forget the fact that autonomous driving is not possible without reestablishment of existing road infrastructure to come up with such necessities as detailed prior map, etc.”