基于多视角视频监督的高保真度仿真环境建立系统及方法
By establishing a robotic arm training simulation environment through multi-view video supervision, and combining semantic segmentation and physical parameter updates, the problem of inaccurate simulation of physical properties of the simulation environment in existing technologies is solved, realizing the rapid establishment of a high-fidelity simulation environment and the high applicability of intelligent algorithms.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- 娄浩哲
- Filing Date
- 2024-05-23
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for establishing training simulation environments for robotic arms cannot accurately simulate the physical properties of real-world environments, resulting in poor applicability of trained intelligent algorithms in complex environments. Furthermore, existing methods are time-consuming or difficult to measure physical parameters.
A multi-view video supervision method is adopted, which records the movement process of the robotic arm through a camera array. Combined with semantic segmentation and physical parameter updates, a high-fidelity simulation environment is established, including RGB texture model, 3D mesh model and URDF file generation, and the physical parameters of the simulation environment are optimized.
It achieves a high-fidelity simulation environment that closely approximates the physical properties of the real environment, reducing setup costs and time, and improving the applicability of intelligent algorithms in complex environments.
Smart Images
Figure CN118596137B_ABST