Update Drone-Env with Random Target and Improved Documentation

[KafuuChikai]

Demo:

video.mp4

Description:

This pull request updates the hover environment to include random target point generation and improves the documentation to reflect these changes. The drone is now trained to reach randomly generated target points, enhancing the training process and making it more dynamic.

Changes:

1. Hover Environment Update (hover_env.py):

   • Updated the hover environment to generate random target points that the drone must reach.
   • Improved the reward functions to provide feedback based on the drone's performance in reaching the target points.
   • Enhanced the visualization.

2. Training Script Update (hover_train.py):

   • Updated the training script to reflect the changes in the hover environment.
   • Ensured that the training process correctly initializes the environment and starts the training with the new random target point generation feature.

3. Evaluation Script Update (hover_eval.py):

   • Updated the evaluation script to reflect the changes in the hover environment.
   • Add target point visualization and record videos.

4. Documentation Update (README.md):

   • Updated the installation and training instructions to ensure clarity and completeness.
   • Added a reference to the inspiration for the reward design.

How to Test:

1. Training:

   • Run the training script with the following command:

     python examples/drone/hover_train.py -e drone-hovering -B 8192 --max_iterations 500

   • The script will initialize the environment, start the training process, and save the trained policy and configurations in the drone-hovering directory.

2. Evaluation:

   • Run the evaluation script with the following command:

     python examples/drone/hover_eval.py -e drone-hovering --ckpt 500 --record

   • The script will load the trained policy from the specified checkpoint, initialize the environment, and visualize the drone's performance.

---

And I'll soon write some tutorials on Genesis World Documentation if that's okay with you.

Please review the changes and provide feedback. Thank you!

[zhouxian]

nice :)