Kakao Mobility said on the 24th that it will take part in the Kakao developer conference "if(kakao)25" to present the company's artificial intelligence (AI) technology achievements.
At this conference, Kakao Mobility will give a presentation focusing on the current state of AI technology development and real-world applications in autonomous driving and navigation. It will also share innovation cases such as "digital twin map-making automation" and "Kakao T Quick AI auto paste for addresses."
First, Park Chang-su, tech leader of the autonomous driving development team, will join the "technology deep-dive session day" on the 24th to deliver a presentation titled "Applying autonomous driving AI to real vehicles: the autonomous driving we are building for service." The core technologies for autonomous driving services will be explained around three pillars: ◇vehicle systems ◇AI frameworks ◇control systems. Beyond driving technologies for commercialization, the talk will share the overall development process of platform technologies for operating and managing them, as well as technical attempts and experiences for applying them to actual services.
Park will also present how the entire process has been systematized for autonomous driving services, from large-scale data collection and processing to building AI learning pipelines and verifying and deploying algorithms, along with the status of building a service platform that enables integrated operation of autonomous vehicles, including dispatch, control, and remote guidance.
Next, Kim Pureummoe, a researcher on the AI research and development team, will introduce a case of adopting reinforcement learning based on "multi-armed bandit (MaB)" for KakaoNavi's real-time route search. This is a methodology that converts the degree of mismatch between the information the navigation considers when searching for routes and the real-world driving environment into an optimized figure and reflects it in large-scale real-time route search.
Kakao Mobility calculated a road-by-road "route compliance rate" based on whether drivers actually drove on the guided road, and used it to assess each road's "travel value," which was reflected in KakaoNavi's real-time route search. For example, if drivers tend not to drive on roads around bus transfer centers where travel is inconvenient, or on roads congested with vehicles that are parked or stopped at all times, the system judges that those roads are likely unsuitable for driving and will avoid guiding routes through them in the future.
On the final day, the 25th, Kakao Mobility will join the "technology asset booth (demo station)," a venue for technology exchange and networking among Kakao Group developers, to showcase various AI adoption cases, including "automated digital twin map-making implemented with AI" and a "Kakao T Quick AI address autocomplete experience."
Ryu Geung-seon, CEO of Kakao Mobility, said, "The technologies introduced at this if(kakao) have gone beyond simple research and are being applied to actual services, improving user experience," adding, "We will continue to provide a smarter and safer mobility experience through innovative AI-based mobility technologies."