Path Planning Using Deep Learning

Data

data_2.zip and data_3.zip contains 2D and 3D data used for training and testing, respectively.

Trained Models

results_2.zip and results_3.zip folder contains learned network models for 2D and 3D case trained on different grid sizes, respectively.

Source Code

src.zip contains Python scripts for training (train_path_...) and testing (predict_path_...) network models for 2D and 3D case, respectively. After training the model will be saved in HF5 file format which later can be used for the path prediction.

Path Planning Examples

2D path planning: Examples of network predictions for single- and multi-path planning. Networks were trained and tested on the same grid size. Note that in case of multi-path planning the network was only trained on single paths.

3D path planning: Examples of network predictions for single-path planning. Networks were trained and tested on the same grid size.

Note that below we only show nine examples for each case whereas each ZIP file contains 2000 and 1000 examples for a single- path and multi-path planning, respectively.

2D Maze:

maze_10x10
path path path
path path path
path path path

maze_15x15

path path path
path path path
path path path

maze_15x15_multi

path path path
path path path
path path path

maze_20x20

path path path
path path path
path path path

maze_30x30

path path path
path path path
path path path

3D Maze:

maze_10x10x10
path path path
path path path
path path path

maze_15x15x15

path path path
path path path
path path path

maze_20x20x20

path path path
path path path
path path path

Computational Neuroscience Group