search

rainorangelemon / gnn-motion-planning

The official repo for NeurIPS 2021 paper 'Reducing Collision Checking for Sampling-Based Motion Planning Using Graph Neural Networks'
MIT License
90 stars 17 forks source link

Dataset about Maze3d #7

Closed wenjinzhang closed 1 year ago

wenjinzhang commented 1 year ago

Hi

I tested GNN explorer with pre-trained weight data/weights/weights_maze_3.pt for Maze 3D environment (map file is maze_files/mazes_hard_3.npz) but the collision check time is as high as 30213.90. Is this because the weight is not well-trained?

Would you please upload the training dataset for the maze 3D environment named data/pkl/maze_prm_3.pkl or update the pre-trained model? Thank you very much!

Best regards, Wenjin

wenjinzhang commented 1 year ago

It seems I can generate a training dataset using algorithm/dijkstra.py by myself. Am I right?