Mars Auto Launches Korea’s First Trailer Autonomous Driving

Full-Scale Establishment of Export Logistics Network to Busan Port
Pursuing Complete Unmanned Freight Transportation Between Korea and the U.S.
Level 4 Autonomous Driving Extended to General Roads

Mars Auto, a startup specializing in autonomous driving for heavy-duty trucks, held a press conference at Dream in Gangnam-gu, Seoul on July 1 and unveiled its vision to achieve full unmanned cargo transportation from Korea to the United States using E2E (End-to-End) AI-based trailer autonomous driving technology.

Mars Auto is a startup that is driving the commercialization of long-haul freight transportation using camera-based E2E AI autonomous trucks. As of the first half of this year, the company has achieved a cumulative autonomous driving distance of 2.5 million kilometers, has been operating 11 paid autonomous freight routes in Korea and the U.S. since 2023, generated KRW 2.5 billion in annual recurring B2B revenue (ARR), and initiated transcontinental autonomous cargo transport in the U.S. exceeding a round-trip distance of 7,000 kilometers. Notably, the company has secured 20 million kilometers of actual truck driving data, which is essential for the advancement of E2E AI, demonstrating its rapid growth in both domestic and international markets.

At the press conference, Park Ilsu, CEO of Mars Auto, and Noh Jaekyung, Vice President, gave a presentation on the theme of ‘Real Self-Driving’. They announced plans to establish an ‘autonomous logistics network’ that will, for the first time in Korea, connect export cargo to Busan Port via trailer autonomous driving. They also officially unveiled Mars Auto’s next-generation Level 4 autonomous driving AI model for heavy-duty trucks, ‘MarsNet 3’, and the subscription-based Level 2 autonomous driving solution for heavy-duty trucks, ‘Copilot’.

Park Ilsu, CEO of Masauto, is giving a presentation on the theme of 'Real Self-Driving' at a press conference on the 1st. Photo by Lee Seungjin

First Trailer Autonomous Driving in Korea... Autonomous Export Logistics Network to Busan Port

Mars Auto has secured a special mobility demonstration exemption from the Ministry of Land, Infrastructure and Transport and recently passed the evaluation by the Korea Transportation Safety Authority. The company plans to officially launch Korea’s first trailer autonomous driving transportation service in the third quarter of this year.

Initially, the company will begin paid trailer autonomous driving services to and from Busan Port in partnership with three clients. Autonomous driving trailers based on Hyundai Motor’s Xcient will be deployed, and since approximately 80% of these routes overlap on common trunk lines, the service can be expanded into a ‘self-driving export cargo network’ utilizing repeated operations. This marks the first application of the ‘Team Korea’ technology and operational know-how, proven through transcontinental autonomous cargo transport in the U.S., to the domestic export logistics field.

On this day, Mars Auto unveiled ‘MarsNet 3’, a heavy-duty truck autonomous driving E2E AI model that extends driving coverage to general roads. MarsNet 3 enables driving not only on highways—which account for 98% of long-haul freight routes—but also on the remaining 2% of general roads, all without HD maps, using a camera-centric system.

The core of E2E AI’s autonomous driving performance lies in real-world driving data and GPU infrastructure. Mars Auto has recently resolved bottlenecks in both real-world data and GPU infrastructure for heavy-duty trucks, accelerating the development of MarsNet 3. By the first half of this year, the company had accumulated 20 million kilometers of real-world heavy-duty truck driving data. In addition, Mars Auto was selected for the national AI project ‘SDV Transition and Advancement of AI Future Vehicle E2E Autonomous Driving Model’, securing exaflops-level NVIDIA Blackwell GPU infrastructure—a key achievement.

In particular, the real-world driving data used for training consists of data collected on fixed paid transport routes as well as 365-day real-world driving data—covering both day and night and adverse weather conditions—gathered nationwide via data collection devices installed on partner trucks. MarsNet 3, currently under development, aims to reach target performance by the end of this year and will undergo beta testing on U.S. freight transportation routes.

MasAuto CI. MasAuto

Launch of Level 2 ‘Copilot’ for Heavy-Duty Trucks... Subscription-Based Autonomous Driving to Secure Data and Revenue in Both the U.S. and Korea

Mars Auto also unveiled ‘Copilot’, a Level 2 autonomous driving service for heavy-duty truck drivers. Copilot is a subscription-based autonomous driving service focused on supporting long-distance operations on highways and trunk roads.

While driver assistance functions are rapidly becoming mainstream in the passenger car market, the heavy-duty truck segment requires more advanced technology due to larger vehicle size, heavier loads, trailer connections, and long-distance operation. As a result, the market is still in its early stages. In North America, the adoption rate of driving assistance features in new passenger cars exceeds 30%, but for heavy-duty trucks, the rate is still below 1%.

Copilot reduces fatigue for long-haul truck drivers who may drive up to 11 hours per day, and improves fuel efficiency through AI-optimized driving. Notably, pilot operations with domestic and international partners have demonstrated fuel cost savings of over 10%, which translates to around KRW 2 million in monthly savings per vehicle for long-haul operations in the U.S. In Korea, Copilot will serve as a foundational technology to enhance the safety and efficiency of highway logistics, in conjunction with the Korea Expressway Corporation's ‘Nighttime Autonomous Truck-Only Lane Pilot Project’.

Furthermore, Copilot serves as a ‘data flywheel’ that accelerates the commercialization of Level 4 autonomous driving. The company plans to continually improve the performance of its Level 4 heavy-duty truck E2E AI by leveraging both stable subscription-based revenue from commercial services and the accumulation of large-scale real-world driving data collected from actual transportation operations.

Park Ilsu, CEO of Mars Auto, stated, “True autonomous driving, as pursued by Mars Auto, goes beyond existing approaches that cost hundreds of millions of won per vehicle. Our goal is to create an economically viable and safe cargo transportation ecosystem through scalable E2E AI technology in real-world logistics operations. We aim to expand trailer autonomous driving in Korea and the U.S., and by advancing AI through MarsNet 3 and the data flywheel, we will achieve full unmanned long-haul ‘middle-mile’ freight transportation by 2028.”

This content was produced with the assistance of AI translation services.