Path-Transformer

Introductionn

This is the open-source repo for our thesis -- Path Transformer: Generating Partial Reference Paths for Smooth Movement in Local Obstacle Avoidance

Quick Start

The project is divided into three sections, namely experience generating, model training, and model testing, in sequential order.

Experience Generating

Start generating:

cd ${root directory}/exp_gen/PRM+A*
# if you are running for the first time
catkin_make --only-pkg-with-deps img_env
# start generating
source devel/setup.bash
roslaunch img_env gen_exp.launch
# open up a new terminal
source devel/setup.bash
python env_test.py

Stop: Cease the Python terminal by using Ctrl+C.

Output: You will find the newly collected experience pool in ${root directory}/exp_gen/PRM+A*/output/.

Model Training:

Put the newly collected experience pool in ${root directory}/train_model/ using methods such as copying or creating a symbolic link. Execute the Python script:

python train.py

Output: You will find the training logs and saved models in ${root directory}/train_model/output/.

Model Test:

Put the newly trained model in ${root directory}/test_model/DT_test/model/ using methods such as copying or creating a symbolic link.

Start testing:

cd ${root directory}/test_model/DT_test
# if you are running for the first time
catkin_make --only-pkg-with-deps img_env
# start testing
source devel/setup.bash
roslaunch img_env DT_test.launch
# open up a new terminal
source devel/setup.bash
python env_test.py

Output: Please be patient and you will see the test results in the Python terminal.