基于单张俯瞰图的 HoloOcean 近岸仿真环境自动化生成管线

Summary of the Thoughts:

Alright, let's break this down. It seems we have a single paragraph here. My initial thought is that marine robotic simulations are fantastic for developing and testing algorithms for UUVs and USVs. They provide a safe and inexpensive way to do this before we commit to full field deployment, focusing on autonomy and perception. The thing is, the simulations are often limited by the simulated environments available. Currently, the marine robotics simulation suites offer manual ways to create or edit these environments, but those methods require either pre-existing data or specialized knowledge of the system. That's a bottleneck, we need to streamline that process. So, to address this, I'm proposing a new Unreal Engine 5 level generation pipeline that will automatically create coastal environments specifically for HoloOcean. The idea is that the pipeline uses a user-provided overhead image of a coastal scene. It then takes that image, generates the height map data from it, and automatically selects and places the assets within the environment. This should significantly speed up the whole process.

海洋机器人仿真提供了一种安全且低成本的方法，用于在全面实地部署前开发和测试无人水下航行器（UUV）与无人水面艇（USV）的自主及感知算法。然而，此类仿真往往受限于仿真环境的可用性。当前的海洋机器人仿真套件虽然提供了手动编辑或创建环境的方法，但通常需要现有数据或对环境系统的专业知识。为了解决这些问题，我们提出了一种新颖的虚幻引擎 5（Unreal Engine 5）关卡生成管线，能够为 HoloOcean 自动创建海岸环境。我们的管线依赖于用户提供的海岸场景俯视图像，进而利用该图像生成高度图数据，并自动选择资产放置在环境中。