Autonomous-driving development often requires teams to assemble vehicles, sensors, computing hardware, software and simulation infrastructure before they can even begin testing an AI system. TIER IV is attempting to lower that barrier with an open reference design for autonomous racing karts, giving developers a common hardware and software foundation built around its Autoware autonomous-driving platform.
TIER IV has released the software and design information behind its autonomous racing kart reference design through a public GitHub repository, combining open-source autonomous-driving software with simulation and vehicle-testing environments.
The platform was used during the 2026 Autonomous Driving AI Challenge, organized by the Society of Automotive Engineers of Japan. According to TIER IV, 240 teams involving approximately 600 participants used the reference design during the competition’s qualifying and final rounds, which included both simulated and real-world driving challenges.
The significance of the project is less about racing itself than the development infrastructure underneath it.
Building an autonomous vehicle from scratch requires integration across sensors, vehicle control, onboard computing, perception, localization, planning and simulation. Developers also need a way to test algorithms repeatedly before putting them onto physical vehicles. For newcomers, those requirements can consume significant resources before any autonomous-driving research begins.
TIER IV’s reference design provides a starting point. Its hardware configuration includes a TOM’S electric racing kart, GNSS, an inertial measurement unit and onboard computing hardware. The software stack is built on Autoware, the company’s open-source autonomous-driving platform.
A publicly available simulator reproduces the racing-kart environment, while an online evaluation system allows teams to test software and review driving results. The GitHub repository contains the autonomous-driving software, simulator and associated design information.
That combination creates an approach increasingly familiar across AI development: standardize the underlying infrastructure while allowing developers to compete on algorithms and models.
The same principle is visible in broader AI ecosystems, where open frameworks and common compute environments allow researchers and developers to build specialized applications without recreating foundational infrastructure. In autonomous driving, however, the problem is more complicated because software must interact with physical sensors, vehicles and safety-critical systems.
The project also gives Autoware a practical development pathway. Rather than limiting the open-source platform to research environments, the racing reference design connects its software with simulation and physical vehicle testing.
For autonomous-driving developers and educational institutions, that can reduce the infrastructure burden associated with experimentation. For enterprises, the more important question is whether similar open reference architectures can shorten development cycles while still meeting the validation, safety and reliability requirements of commercial autonomous systems.
TIER IV says it plans to expand its reference designs and development environments while working with industry, research and education partners. The broader objective is to grow an open ecosystem around autonomous driving rather than keeping development infrastructure inside individual technology organizations.
Market Landscape
Autonomous-driving AI depends on a technology stack that spans computer vision, sensor fusion, localization, machine learning, simulation, edge computing, vehicle control and safety engineering.
Open-source frameworks such as Autoware provide developers with reusable software components, while ecosystems from companies such as NVIDIA, Qualcomm, Mobileye and Waymo represent other approaches to autonomous-driving compute and software infrastructure.
Simulation is becoming particularly important because AI-based driving systems require enormous numbers of test scenarios. Developers need to evaluate perception and planning systems under different road layouts, weather conditions, traffic patterns and edge cases before deploying them on public roads.
TIER IV’s racing-kart reference design addresses a narrower environment, but its architecture demonstrates how standardized hardware, software and evaluation infrastructure can make physical AI development more accessible.
The key enterprise challenge remains validation. An autonomous-driving platform must demonstrate not only that an AI model works in simulation, but that the complete vehicle system behaves predictably when exposed to real-world conditions.
Top Insights
- TIER IV has released an open autonomous-racing reference design combining Autoware software, vehicle hardware, simulation and evaluation tools.
- The platform was used by 240 teams and approximately 600 participants during Japan’s 2026 Autonomous Driving AI Challenge.
- Public software and design resources allow developers to focus on autonomous-driving algorithms instead of rebuilding complete vehicle-development infrastructure.
- Simulation and real-world testing are integrated into the development workflow, creating a common environment for evaluating autonomous-driving software.
- The project illustrates how open-source infrastructure can lower barriers to physical AI development while creating new pathways for autonomous-driving talent.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
