English 
SC Automotive Engineering Co., Ltd., a company specializing in development engineering for automobiles, has signed a sales agent and distributor agreement with Lightwheel.
In recent years, in the fields of robotics and autonomous driving, AI models used for recognition, action planning, control, etc. have become increasingly sophisticated. Training and evaluating these AI models requires vast amounts of diverse data that cannot be collected in the real world alone. Lightwheel offers the following solutions as an AI learning platform that utilizes advanced simulation technology.
Embodied AI (Robotics) Solutions
In the field of robotics, there is an overwhelming lack of action data for AI training, and while data collection is currently being carried out through imitation learning, where humans teach robots how to move, there are limitations to collecting data in the real world due to resource and time constraints.Lightwheel enables efficient robot training and evaluation by providing high-precision SimReady assets (high visual and physical fidelity, ready to use in simulation) that enable learning in a simulated space, EgoSuite (first-person human action data for basic model training), and an evaluation platform.
こちらもお読みください： SiMa.ai、シノプシスと自動車向けフィジカルAIで提携
Autonomous driving solutions
In recent years, advanced AI models such as E2E (End-to-End) and VLA (Vision-Language-Action) have begun to be introduced in the autonomous driving field, increasing the need for vision-based training and evaluation data for various scenarios. However, it is difficult to collect real-world data for dangerous or rare scenarios, or for the same scenarios where only the weather or the behavior of other vehicles differ. Lightwheel recreates various scenarios in a realistic simulation space using 3D reconstruction from real-world data and synthetic data, and generates training and evaluation data through sensor simulation, enabling efficient training and evaluation of autonomous driving AI.
ソース PRタイムズ
日本語