Associated code for our paper, Adaptive Online Planning for Continual Lifelong Learning. See our website for more details.
- Clone/download a copy of this repository.
- Code uses Python 3, as well as the following packages, which can be installed via pip/conda: numpy, gym, scipy, torch, matplotlib, and seaborn.
- Install MuJoCo and mujoco-py.
To run an experiment, run the command (all args are optional, use -h for help/more information):
python run.py --a aop -e hopper -s changing
To visualize results, identify the directory of the experiment and run (replace ex/1124_1200 with relevant directory and 20000 with the length of the experiment):
python graph.py ex/1124_1200 20000