https://github.com/TianxingChen/RoboTwin/tree/main?tab=readme-ov-file

According to the readme, first configure the environment. I still use the dl environment.

Then check the config according to readme.

For better visualization, let’s use 20 and true. Let’s see the effect.

In this code configuration, fovy refers to Vertical Field of View.

It indicates the angle range of the scene that the camera can capture in the vertical direction, usually measured in degrees (°).

• For the L515 camera, fovy: 45 means its vertical field of view is 45 degrees.

• For the D435 camera, fovy: 37 means its vertical field of view is 37 degrees.

This parameter is related to the camera’s focal length and sensor size, and determines how wide the “up and down” view of the camera is. Correspondingly, there is the horizontal field of view angle (fovx or hfov), which represents the “left and right” viewing range.

But when using eval, I found that this thing couldn’t even lift a hammer. I laughed.

First, create a gap with a clamp, then move around nearby, proving that reinforcement learning still has issues on agile.x.