Until recently, only a small number of companies with expertise in high-priced specialized sensors and the automobile industry could implement autonomous navigation technology. High technological and financial barriers formerly allowed only ICT giants and a small number of companies in the automobile industry with capabilities for long-term investment and technology development to take the initiative in the development of autonomous driving technology.

In fact, a single car used in the early days of Google’s development of autonomous driving was priced at about 170 million won. The price of a LIDAR sensor used as the core sensor of the car stood at 80 million won. In addition, about 170 staff members brought in mainly from automakers spent more than four years in R&D to develop a self-driving vehicle.

However, these high technological barriers are being broken down by artificial intelligence, especially deep learning. Companies that implement self-driving technology using deep- learning have been emerging rapidly over the past two years, particularly in Silicon Valley.

Unlike previous autonomous driving technology, which was implemented by automobile experts using a rule-based approach, these companies implement autonomous driving technology in a process as if people improved driving skills by repeating drives based on deep learning. A small number of developers are implementing autonomous driving technology in a very short time by using low-cost, general-purpose sensors rather than expensive ones.

In addition, the rapidly evolving artificial intelligence technology is expected to be applied to autonomous navigation and will accelerate the revolutionary change of technology competition in the future. Leading research institutes in the field of autonomous navigation are already conducting studies to integrate the latest research in artificial intelligence such as reinforcement learning, relational networks, and transferring intelligence into the development of autonomous navigation technology.

Especially, as these researches realize a process where artificial intelligence learns, reasons, and forecasts in a way similar to that of human beings, when applied to autonomous driving technology, it will be possible to develop a car that runs by thinking and judging like a human being.

In academia and the start-up world, autonomous driving technologies based on deep learning have already been introduced one after another. With innovative cases taking place every year, experts in the fields of artificial intelligence and deep learning, not experts in the automotive industry, are integrating their research into the automotive field and presenting their technology.

In particular, these companies are announcing their technologies as open sources and they are accelerating technology through researchers’ participation and competition. This method differs greatly from technology competition among major companies in the automobile industry which have developed technology through the development and internalization of their own technology.

Automakers that have adhered to existing methods are also responding to the new technology paradigm by rapidly securing their deep-learning capabilities. Major automakers with OEM systems such as Daimler, VW, and Toyota have been rapidly introducing external technologies by investing in and acquiring start-ups related to deep learning since 2016. Internally, they have also invested heavily in developing their own technologies by setting up AI-specialized research centers. In particular, GM and Ford are responding to technological competition where they have been lagging behind their rivals by acquiring or investing in self driving start-ups based on deep learning with more than one trillion won in investments each.

“The core of autonomous technology competition has already started to move to deep learning. Artificial intelligence experts in the field of deep learning are rapidly implementing autonomous navigation technology with low-cost general-purpose sensors. Most companies are also developing autonomous navigation technology based on deep learning and competition in the automobile industry is expected to focus on securing competence in artificial intelligence field and securing driving data in the future,” said Lee Seung-hoon, a researcher at the LG Economic Research Institute. “This is because the technological quality of AI based on machine learning such as deep learning will be determined by data collected more diverse driving environments and the utilization of learning processes.”

In addition, he said, “Companies such as comma.ai and Tesla that have already recognized the importance of securing actual driving data have already gathered driving data from millions to hundreds of millions of kilometers and are using them in self driving learning processes. It is expected that when the autonomous driving market is full-fledged, technology gaps between companies with such data and advanced intelligence and those without them in the beginning of the market will be very large and it will be a big challenge for latecomers to narrow such gaps,” Lee added.

korean-machinery.com | Blog Magazine of korean-machinery, brands and Goods