MasAuto, a startup specializing in autonomous driving for large trucks, has announced its vision to deliver South Korean export goods to the United States using fully unmanned technology.
Utilizing end-to-end (E2E) AI-based trailer autonomous driving technology, the company aims to achieve 100% unmanned delivery from Busan Port in South Korea to West Point in the U.S., via Long Beach Port, while also addressing logistics efficiency through 'empty transport.' The company reports a surge in orders from clients in South Korea and the U.S., projecting revenues of 6.3 billion won this year and 14.9 billion won by 2027.
During a press briefing on July 1 in Gangnam, Seoul, CEO Park Il-soo stated, "The biggest issue with existing autonomous driving systems is the high initial investment cost and the difficulty of technological scalability. To address this, MasAuto's 'Mas Pilot' offers autonomous driving technology that can be deployed immediately at a cost that is one-twentieth of traditional methods, providing limitless scalability."
Founded in 2017 by Park, a KAIST graduate, MasAuto has developed a camera-based E2E AI similar to Tesla's, with the goal of achieving complete unmanned freight transport by 2028.
The flagship product, 'Mas Pilot,' is already installed in 14 domestic clients' fleets, operating autonomously on 97% of their routes. As of the first half of this year, the technology has accumulated 2.5 million kilometers of autonomous driving, successfully gathering 20 million kilometers of real-world driving data essential for enhancing E2E AI. The company has also recorded autonomous freight transport over distances exceeding 7,000 kilometers in the U.S.
At the briefing, MasAuto announced plans to establish a 'self-driving logistics network' that will connect export goods to Busan Port, marking the first introduction of trailer autonomous driving in South Korea. The company also unveiled its next-generation Level 4 autonomous driving AI model, 'MarsNet 3,' and a subscription-based Level 2 autonomous driving solution for large trucks called 'CoPilot.'
MasAuto recently secured a mobility demonstration exemption from the Ministry of Land, Infrastructure and Transport and passed evaluations by the Korea Transportation Safety Authority, allowing it to begin the country's first trailer autonomous transport in the third quarter. Trailers account for over 95% of South Korea's export container logistics, and since more than 60% of overseas export volumes are processed through Busan Port, this route establishment is critical for enhancing the competitiveness of domestic export logistics.
Park stated, "We plan to start paid autonomous transport with three clients using Hyundai Xcient trucks between Busan Port and the designated routes. This route includes a common corridor where approximately 80% of the segments overlap, allowing for the expansion of an 'export freight autonomous driving network' based on repeated operations." This marks the first application of the technology and operational know-how validated through the 'Team Korea' initiative for cross-country autonomous freight transport in the U.S. to domestic export logistics.
The E2E AI model 'MarsNet 3' expands its driving capabilities beyond highways to include public roads. While 98% of long-distance freight transport occurs on highways, about 2% requires transport on public roads. The key is navigating complex urban environments using camera-based systems without relying on high-definition maps.
To enhance the autonomous driving performance of E2E AI, real-world driving data and GPU infrastructure are essential. MasAuto's Vice President No Je-kyung explained, "To accelerate the development of MarsNet 3, we accumulated 20 million kilometers of real-world driving data in the first half of this year and secured exaflops-level NVIDIA Blackwell GPU infrastructure through selection for the national AI project 'SDV Transition and AI Future Vehicle E2E Autonomous Driving Model Advancement.'"
He added, "We will enhance the learning data based on fixed routes for paid transport across the national road network, supplemented by 365 days of real-world driving data collected from partner trucks in various conditions, including day and night and adverse weather. We plan to conduct beta testing on U.S. freight transport routes after achieving target performance by the end of this year."
The subscription service 'CoPilot' for Level 2 autonomous driving for long-haul truck drivers was also introduced. CoPilot aims to reduce fatigue for drivers operating up to 11 hours a day and improve fuel efficiency through AI-optimized driving. The service has demonstrated over 10% fuel cost savings through pilot operations with domestic and international partners, translating to approximately 2 million won in monthly savings per truck based on U.S. long-haul transport standards. It will be utilized as a foundational technology to enhance the safety and efficiency of highway logistics in conjunction with the Korea Expressway Corporation's pilot project for dedicated lanes for autonomous trucks during nighttime hours.
Additionally, CoPilot will serve as a 'data flywheel' to accelerate the commercialization of Level 4 autonomous driving. By generating stable subscription revenue through commercial services, the company aims to continuously improve the performance of large truck Level 4 autonomous driving E2E AI based on the extensive real-world driving data accumulated in actual transport operations.
Park emphasized, "True autonomous driving goes beyond the existing methods that cost hundreds of thousands of dollars per vehicle. We aim to create an economically viable and safe freight transport ecosystem with scalable E2E AI technology in logistics. We plan to expand trailer autonomous driving in both South Korea and the U.S. and enhance AI capabilities based on MarsNet 3 and the data flywheel to achieve complete unmanned long-distance freight transport by 2028."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.