[Edaily Reporter KIM JI-WAN ] INNOSIMULATION CO., LTD.(274400)has unveiled verification technology that translates AI decisions made in a virtual space into actual movements of autonomous vehicles. This “Physical Twin” technology links simulations with real equipment to verify that the AI operates as intended in the real world.

INNOSIMULATION CO., LTD. announced on the 5th that it is participating in “2026 RoboWorld,” taking place at KINTEX Hall 1 in Goyang, Gyeonggi Province, through the 7th, to introduce its Physical Twin technology and the “INNO ALPS” Physical AI verification platform.

The highlight of this exhibition is a demonstration of AI-powered autonomous driving using NVIDIA’s “Alpamayo.” The virtual driving environment created by the company’s proprietary automotive simulator, “iMDS,” and sensor data have been integrated with an actual Unmanned Ground Vehicle (UGV). The system works by having the AI assess virtual road conditions and issue driving commands, which the actual UGV then executes accordingly.\\

Visitors can observe both the method where a human remotely controls the UGV and the method where the AI independently makes driving decisions based on sensor data. According to the company, visits from domestic and international corporate representatives and visitors have continued since the opening day on the 4th, and inquiries regarding technology adoption and potential collaborations have also been received.

According to NVIDIA’s official technical documentation, AlphaMayo consists of AI models, simulation tools, and datasets that support the development of inference-based autonomous driving. Among these, the AlphaMayo 1 model is designed to predict a vehicle’s driving path while presenting the rationale behind those decisions in a human-readable format.

INNOSIMULATION CO., LTD. also unveiled robot arm verification technology developed in collaboration with Vectioneer, a Dutch robotics and mechatronics company. This technology creates an environment where virtual objects interact with a robotic arm to test the robot’s movements and its ability to perform tasks. Vectioneer, a motion control specialist founded in 2010, owns the real-time control platform “MOTORCORTEX.”

The challenge addressed by these two demonstrations is the discrepancy between the virtual environment and reality. Even if the AI of a robot or autonomous vehicle operates normally in a simulation, it may produce different results on actual equipment due to sensor errors, friction, and drive system characteristics. This is referred to as the “Sim-to-Real Gap.”

A physical twin is a validation environment designed to bridge this gap by connecting a digital twin to actual hardware, sensor data, a motion platform that replicates movement, and a dynamometer that measures drive performance. In particular, by utilizing “VIL (Vehicle-in-the-Loop)” technology—which integrates actual vehicles, sensors, and drive systems into a virtual environment—the company explains that even situations that are dangerous or difficult to replicate on-site can be repeatedly tested under controlled conditions.

INNOSIMULATION CO., LTD. plans to use this approach to verify the operational performance of AI trained in a virtual environment on actual equipment and to secure the test data necessary for industrial application.

INNOSIMULATION CO., LTD. stated, “At this exhibition, visitors are showing continued interest in the physical twin demonstration that connects virtual simulation with an actual UGV,” adding, “For physical AI to be applied in real industrial settings, it is crucial to verify not only the AI’s decision-making capabilities but also how accurately and safely it operates on actual hardware.”

The representative continued, “Based on 26 years of accumulated expertise in simulation and hardware integration, INNOSIMULATION CO., LTD. is building an infrastructure that connects the entire process from training to verification of Physical AI,” and added, “We plan to continue expanding our Physical Twin technology and platform so that AI can be safely applied in real-world environments across various industries.”

Meanwhile, the company will host a seminar on physical twin technology at 3:00 p.m. on the 5th in the conference room located in Halls 3–5 of KINTEX Exhibition Hall 1. Terence Fung, Head of the Robotics AI Team, will deliver a presentation in “Track C” (the mobility track), introducing the concept and operating principles of physical twins, as well as verification methods that bridge the gap between simulation and actual hardware.