Repository for code and videos from "Bottom-Up Meta-Policy Search", submitted as workshop paper.
In this repository, we have five folders:
- imitation-learning: Data from keyframe initial policy and scripts to train a policy from this dataset
- BUMPS: Files related to BUMPS meta-training, including expert policies dataset, scripts and meta-policy config files
- baselines (PPO1): Code from OpenAI Baselines PPO algorithm used to train single-task expert policies and multi-task RL experiment.
- bootstrap_ci_scripts: Scripts for computing boostrapped confidence interval and plot it.
- video: Folder with illustrative video of kick evaluation.