Kakao Mobility Unveils Level 4 Autonomous Driving Roadmap for Physical AI

Kakao Mobility is making a strategic push into fully autonomous driving technology, aiming to develop Level 4 capabilities in-house as a cornerstone of its broader “Physical AI” initiative. This ambitious move signals a significant investment in shaping the future of mobility, moving beyond incremental advancements to foundational technology development.

The company’s vision was laid out by Kim Jin-kyu, vice president and head of Kakao Mobility’s Physical AI division, at the recent 2026 World IT Show conference in Seoul. His presentation, centered on the integration of autonomous driving into sophisticated mobility platforms within the evolving Physical AI landscape, underscored the company’s commitment to a tangible, real-world application of artificial intelligence.

The conference itself, themed “Beyond Idea, Into Action: AI moves Reality,” brought together over 460 entities from 17 nations, highlighting a global convergence of interest in AI’s practical implementation. South Korea’s Ministry of Science and ICT has actively championed this transition, emphasizing the application of AI to tangible industrial sectors, a sentiment that Kakao Mobility’s strategy directly reflects.

Kim articulated Kakao Mobility’s intent to intricately weave autonomous driving technologies with existing physical infrastructure across Korea. This integration is not merely about deploying vehicles but about fostering a robust, open ecosystem designed to bolster local technological competitiveness and innovation in the autonomous driving space.

Level 4 autonomy, as defined by the U.S. National Highway Traffic Safety Administration, signifies a critical threshold where autonomous systems can manage all driving operations within specific, geofenced operational design domains (ODDs) without requiring any human oversight or intervention. These are typically confined to designated areas, such as autonomous taxi zones or specific business districts, enabling safe and controlled deployment.

Navigating the Level 4 Autonomous Driving Landscape

Kakao Mobility’s roadmap for achieving Level 4 autonomy is strategically segmented into three core technological pillars: advanced machine learning models, robust vehicle redundancy systems, and rigorous validation frameworks.

At the heart of this development lies the creation of sophisticated machine learning models. These are engineered to autonomously handle perception, decision-making, and control functions, eliminating the need for human input. This encompasses the complex task of how an autonomous vehicle interprets its environment, makes critical driving judgments, and executes precise maneuvers.

Furthermore, Kakao Mobility is prioritizing vehicle architectures that incorporate significant redundancy. This design philosophy ensures that critical systems can continue to operate seamlessly even in the event of a component failure, a non-negotiable requirement for safety and reliability in autonomous operations.

The company’s validation platform represents a dual-pronged approach, synergizing high-fidelity virtual simulations with extensive real-world driving data. This comprehensive system is crucial for rigorous testing, continuous performance optimization, and ensuring the highest quality standards as the company refines its autonomous driving services.

Fortifying Safety and Operational Control

Beyond the core driving technology, Kakao Mobility is building a comprehensive safety management platform tailored for autonomous vehicles. A key component of this platform is the Autonomous Vehicle Visualizer, a sophisticated 3D visualization tool designed to provide passengers with a real-time view of the vehicle’s operational field of vision. This feature aims to enhance transparency and passenger confidence by allowing them to actively monitor driving conditions and understand the vehicle’s situational awareness.

The Visualizer is intended to clearly depict what the autonomous vehicle is detecting and processing during its operation, offering passengers an intuitive understanding of the driving context.

Looking ahead, the company plans to augment this with a 24-hour operational control center and an advanced anomaly detection system leveraging cutting-edge vision-language models. These integrated systems are designed to provide real-time situational analysis, enable remote intervention capabilities, and ensure swift emergency response protocols.

The proposed control center will serve as a central hub for monitoring autonomous driving services post-deployment. While Kakao Mobility has indicated the use of vision-language models for anomaly detection, specific details regarding model architecture and performance metrics remain proprietary.

Cultivating an Open Autonomous Driving Ecosystem

A significant aspect of Kakao Mobility’s strategy involves fostering an open ecosystem. The company plans to selectively share key technology assets with a broad range of industry players, including startups and established manufacturers involved in autonomous driving development.

These shared assets are set to include extensive, large-scale autonomous driving datasets crucial for model training, high-definition (HD) maps that provide granular road detail essential for precise localization and navigation, and platform APIs for ride-hailing and dispatch operations. HD maps are a foundational element for autonomous driving, offering the detailed environmental context that vehicles need to make informed driving decisions.

This asset-sharing initiative is designed to accelerate development across the industry, enabling other participants to advance their autonomous driving technologies without the prohibitive cost and time investment of building all foundational infrastructure from scratch.

Kakao Mobility also intends to share operational resources, such as fleet management systems and on-site response capabilities. These contributions are integral to the company’s overarching goal of nurturing a vibrant and collaborative domestic autonomous driving ecosystem.

Insights from Gangnam Autonomous Service

The company highlights its ongoing late-night autonomous vehicle service in Seoul’s affluent Gangnam district as a testament to its current operational capabilities. This service, accessible through the familiar Kakao T platform, allows users to seamlessly integrate autonomous driving options with their existing mobility preferences.

Data from the Seoul Metropolitan Government indicates that the Gangnam late-night autonomous taxi service, operational since September 26, 2024, recorded 7,754 rides through February 28, 2026. Notably, the city reported no accidents attributed to the autonomous driving technology during this period, with the service averaging approximately 24 trips per operational day, demonstrating a promising safety record and operational efficiency.

The service transitioned from a free pilot program to a paid commercial offering in April 2026. Seoul also expanded the operational fleet from an initial three vehicles to seven, supplemented by two reserve vehicles, further solidifying its presence and service capacity.

Accessing the Gangnam service is streamlined through the Kakao T application, where users can select the Seoul Autonomous Car icon or utilize the standard taxi-hailing menu. Kakao T’s integrated approach consolidates multiple mobility services, including taxi booking, navigation, and other vehicle-related services, into a single, user-friendly interface.

The Gangnam service exemplifies how Kakao Mobility is leveraging its established mobility platform to introduce and scale advanced autonomous driving solutions for public use.