A few questions

[yuejunpeng]

Hello, thank you for your work!
I have a few questions:

1. Can create_traj and other codes that interact with the environment use player instead of agent? The agent is a robot, and the viewing angle is relatively low.
2. "Legent launch --scene 1" cannot successfully render the environment.
3. Could you open source the create_traj of the grab task?
4. Could the grab action specify the object ID? That is, use the object ID as an input parameter of grab, and the same applies to the put down action.
5. Unable to use the top-down view.

[chengzl18]

Hello, sorry for the late reply. We are currently working on these issues, and they will all be resolved by October.

1. You can initialize a view similar to the player's by using env = Environment(env_path="auto", use_animation=False, camera_resolution_width=896, camera_resolution_height=448, camera_field_of_view=60). This will set the vertical field of view to 60 degrees, with an image aspect ratio close to 2:1. Additionally, we are developing a module for customizing the agent, including the 3D model and camera position of the agent.

2. We did not include this scene in the latest build. We are organizing more prefab scenes in a better format, which will be released together with the scene information.

3. Yes, we will open source the data generation codes for all task templates, including the diversified data generation codes, in October.

4. This functionality already exists in the environment but hasn't been provided as an API yet. We will include it in future updates.

5. The top-down view has been replaced with a free view and scene editing tool currently under development, which will also include the top-down view functionality. The free view is very useful and we will update the documentation on how to use this free view.