<palign="center">
System Overview
- Ubuntu 20.04
- Python 3.6
The file of the conda environment is environment.yml. We use [V-REP 3.5.0] as the simulation environment.
We do experiments on a NVIDIA 1080 Ti GPU. It requires at least 8GB of GPU memory to run the code.
To train a regular TCG policy from scratch in simulation, first start the simulation environment by running V-REP (navigate to your V-REP directory and run ./vrep.sh
). From the main menu, select File
> Open scene...
, and open the file simulation/simulation.ttt
. Then navigate to this repository in another terminal window and run the following:
To train from scratch, run
python main.py
python main.py
--is_testing --test_preset_cases --test_target_seeking
--load_ckpt --critic_ckpt CRITIC-MODEL-PATH --coordinator_ckpt COORDINATOR-MODEL-PATH
--config_file TEST-CASE-PATH
We use different colors to label the target object in simulation/random
.
To design your own challenging test case in simulation/preset
.
We reference the following code in our project