What Mars Auto Does
Most of the autonomous-trucking industry mapped the world in high definition and layered it with LiDAR. Mars Auto took the road less scanned.
Mars Auto is a South Korean company building self-driving trucks for long-haul freight. Founded in Seoul in 2017 by two computer scientists from KAIST, Ilsu Park and Gyuri Im, it set out to answer a narrow, expensive question: can a truck drive an unfamiliar highway using cameras alone?
Its answer is a "vision-focused end-to-end AI driving agent." Seven onboard cameras feed a single neural network that handles perception, decision-making, and control together, rather than stitching those steps into separate modules. There is no LiDAR and no pre-built high-definition map. CEO Park has likened the approach to Tesla's - let the AI look at the world and decide.
That design choice is also the business case. Because the system reads the road in real time, it can operate on routes it has never seen, and because it leans on cameras rather than costly sensors, it can be retrofitted onto Class 8 trucks that fleets already own. The company reports more than 1.2 million autonomous miles and roughly 20 million kilometers of real truck-driving data collected across Korea and the United States.
Mars Auto is not chasing robotaxis or the delivery-robot last mile. It is aiming squarely at the middle mile - the long, monotonous hub-to-hub hauls where a human driver spends ten hours holding a lane. That is the segment where automation saves the most fuel and the most labor, and where Mars Auto has chosen to prove itself with paying freight instead of perpetual pilots.
How The Truck Sees
End-to-end pipeline · camera in, control out
No HD map is loaded in advance. The same network that watches the road also drives the truck, and every mile driven becomes training data for the next model - a flywheel Mars Auto seeded, unusually, partly inside the video game Euro Truck Simulator before moving to real highways.
How It Is Different
The contrast with the field is the whole pitch: fewer sensors, no maps, and a kit that bolts onto trucks already on the road.
| Approach | Mars Auto | Typical AV-truck stack |
|---|---|---|
| Primary sensor | Cameras (vision-only) | LiDAR + radar + cameras |
| HD maps | Not required | Usually required |
| Architecture | End-to-end neural net | Modular pipeline |
| Vehicle | Retrofit existing trucks | Often purpose-built |
| Focus | Middle-mile long haul | Long haul / mixed |
| Reported fuel gain | 12%+ efficiency | Varies |
The Numbers On The Road
A small team, but a large odometer. Selected figures the company reports, shown to scale.
Bars are scaled for visual comparison, not a common unit. Figures self-reported by Mars Auto and approximate.
Products & Services
Vision-based driving agent
A single end-to-end neural network reads seven cameras and outputs steering, throttle, and braking - no LiDAR, no HD maps, so it can handle roads it has never seen.
Retrofit autonomy kit
Computer, cameras, and actuators that install on existing Class 8 trucks to add supervised self-driving, instead of requiring a new purpose-built vehicle.
Autonomous freight service
Supervised self-driving hauls that move real cargo hub-to-hub for logistics and manufacturing customers in Korea and the U.S.
Fleet data network
A carrier network the company says spans 30+ carriers and 100,000+ daily miles, feeding real-world data back into the model and reporting 12%+ better fuel efficiency.
Business, Customers & Market
Mars Auto's model is business-to-business: it earns revenue by moving freight with supervised autonomous trucks and by deploying its retrofit system for fleet operators and manufacturers who want lower fuel and labor costs. Its wedge is the middle mile - the hub-to-hub segment that is repetitive enough to automate and expensive enough to matter.
The customer list doubles as a proving ground. In Korea, Mars Auto has operated alongside Hyundai Mobis, Hyundai Glovis, CJ Logistics, Lotte Global Logistics, Emart24, Korea Post, and even the Republic of Korea Army. When it hauled Hyundai-exported auto parts across the United States, each run was both a delivery and a demonstration.
Where it fits in the market: Mars Auto is a small, contrarian challenger in a field dominated by better-funded U.S. players. It competes not on the size of its sensor budget but on cost and retrofit-ability - the bet that vision plus AI can do what a $100,000 sensor stack does, for a fraction of the price.
Operating partners & customers
Company File
- Legal name
- Mars Auto, Inc.
- Founded
- 2017, Seoul
- Founders
- Ilsu Park (CEO), Gyuri Im
- HQ
- Seoul, South Korea
- U.S. base
- Round Rock / Austin, Texas
- Team
- ~15-17 people
- Total raised
- ~$12.5M
- Series A
- ~$12M, 2022 (GFT Ventures)
- Seed
- ~$1.5M (Kakao Ventures)
- Accelerator
- Y Combinator, 2019
- Segment
- Middle-mile autonomous freight
Timeline
Founded in Seoul
KAIST computer scientists Ilsu Park and Gyuri Im start Mars Auto to automate long-haul trucking.
Y Combinator & seed round
The company goes through Y Combinator and raises roughly $1.5M, including from Kakao Ventures.
First autonomous pilot in Korea
Mars Auto begins piloting its camera-based system on Korean roads.
Founder on Forbes 30 Under 30 Asia
Ilsu Park is recognized on the Forbes list for industry, manufacturing & energy.
$12M Series A
GFT Ventures leads the round to scale the vision-first autonomy stack.
Commercial freight in Korea
Mars Auto launches the country's first commercial long-haul autonomous freight service under a regulatory sandbox.
U.S. operations begin
Paid freight starts from a Texas base, testing corridors around Dallas, Houston, and San Antonio.
Record haul & U.S. expansion
A ~3,379 km autonomous run from Long Beach to Alabama, plus expanded U.S. operations backed by Korea's industry ministry.
Watch & Explore
FAQ
What does Mars Auto do?
It builds autonomous driving technology for long-haul trucks and operates supervised self-driving freight, aiming to fully automate hub-to-hub shipping.
How is its technology different from other self-driving companies?
It uses a vision-only, end-to-end AI - seven cameras and a single neural network - with no LiDAR and no HD maps, and it can retrofit onto existing trucks.
Who founded Mars Auto and when?
It was founded in 2017 in Seoul by Ilsu Park (CEO) and Gyuri Im, both from KAIST's computer science department.
How much funding has Mars Auto raised?
Roughly $12.5M total, including a ~$1.5M seed (with Kakao Ventures) and a ~$12M Series A led by GFT Ventures in 2022.
Where does Mars Auto operate?
In South Korea, where it ran the first commercial long-haul autonomous freight in 2023, and in the United States from a Texas base since 2024.
Links & Sources
Profile compiled from public sources including Mars Auto, Y Combinator, Forbes, and press coverage. Figures are approximate and self-reported where noted.