A demonstration at the Hanwha Ocean shipyard will test whether a Humanoid Robot can be applied to shipbuilding sites where there are many high-risk and non-routine tasks. Humanoid Robot corporations AeiROBOT and the 3D AI corporations NdotLight plan to train and verify the robot in a virtual shipyard environment, then confirm the feasibility of performing on-site tasks such as heavy-load transport, walking on rough terrain, and obstacle avoidance.
AeiROBOT and NdotLight said on the 8th that they will push a Physical AI-based humanoid demonstration at the Hanwha Ocean Geoje shipyard. The two companies, partners participating in the Nvidia Inception ecosystem, will combine Humanoid Robot and 3D simulation technology to verify applicability to shipbuilding within a year.
Shipbuilding is considered a suitable manufacturing site to examine the potential use of Humanoid Robot. Shipyards have many large blocks, complex work routes, cramped interior spaces of vessels, and high-risk inspection zones, making it difficult to apply existing automation equipment to quite a few tasks. As high-intensity, high-risk work overlaps with a shortage of skilled workers, demand for Humanoid Robot that can assist people or take over some physical tasks is also growing.
In this demonstration, NdotLight will build a foundation for the robot to learn and be verified in a virtual environment before entering the actual site. Using Nvidia Omniverse and Nvidia Isaac Sim, it will create a 3D space similar to a shipyard and provide sim-ready data necessary for robot training. Sim-ready data refers to 3D data containing physical properties and structural information so that robots can recognize objects and learn movements in a virtual environment.
NdotLight will also build a Digital Twin environment similar to the actual shipbuilding site by using its own 3D CAD engine and 3D AI technology. A Digital Twin implements the real world in a virtual space, allowing robots to test various scenarios before being deployed to an actual shipyard. NdotLight plans to generate 3D asset, a physics-based simulation environment, and synthetic data needed for humanoid training and verification.
AeiROBOT will verify whether its own Humanoid Robot Alice can perform the physical tasks required at shipbuilding sites. AeiROBOT plans to use Nvidia KIMODO and SOMA Retargeter to train the Humanoid Robot with motion data from industrial sites. It will then confirm the feasibility of performing real tasks such as heavy-load transport, autonomous movement, walking on rough terrain, obstacle avoidance, and tool manipulation. AeiROBOT's Humanoid Robot Alice was introduced as a Physical AI example performing shipyard welding work during Nvidia CEO Jensen Huang's CES keynote in January.
Through this project, Hanwha Ocean will examine the industrial potential of applying Humanoid Robot to shipbuilding sites. Humanoid Robot can use the same workspaces, equipment, and movement routes that people use, expanding applicability even to areas that are difficult to address with existing automation equipment. Accordingly, Hanwha Ocean plans to look into the use of Humanoid Robot for inspecting hazardous zones, repetitive on-site movement, equipment and goods transport, and assisting workers.