This is a fork of the original repository under the project Smart_POV which funds basics for intelligence development on simulated voyager agents. The original code is editted such that it simply interacts - and may be evaluated - with our developed professional (D)RL-evaluation platform MIIO2V. A comparison of this project with other similar works id documented here.
- Ubuntu 20.04
- ROS Noetic
- Anaconda
- Clone the repository.
git clone https://github.com/spov-research/purl.git
cd purl- Create a conda environment with python 3.6 and tensorflow 1.13.1 then activate it.
conda create -n purl python=3.6 tensorflow=1.13.1
activate purl- Install the requirements.
pip install -r requirements.txtTo train a model, run:
./purl trainFor example, to train a model using the PPO algorithm on the MiniGrid-LavaCrossingS9N1-v0 environment, use the following arguments:
./purl train --algorithm ppo --environment MiniGrid-LavaCrossingS9N1-v0
- Algorithms
MDP (using the
FullyObsWrapper)
- Q-table
- Q-network POMDP
- PPO
- DQN (with the Double DQN extension) - work in progress, not currently working as intended
- DRQN - work in progress, not currently working as intended
To visualize a model, run:
./purl vizualize