Meta-Reinforcement Learning in Non-Stationary and Dynamic Environments

The code of CEMRL is released at https://github.com/BZSROCKETS/cemrl .

All experiments (newly added videos are at the bottom of this page)

all_experiments_new.mp4

Single experiments

Stationary, Parametric Environments

cheetah_stationary_dir.mp4

Half-Cheetah stationary direction

cheetah_stationary_vel.mp4

Half-Cheetah stationary velocity

Non-Stationary, Parametric Environments

cheetah_non_stationary_dir.mp4

Half-Cheetah non-stationary direction

cheetah_non_stationary_vel.mp4

Half-Cheetah non-stationary velocity

ant_non_stationary_dir.mp4

Ant non-stationary direction

ant_non_stationary_vel.mp4

Ant non-stationary velocity

Continual Learning

cheetah_continuous_learning_dir.mp4

Half-Cheetah continual learning

Stationary, Non-Parametric Environments

cheetah_stationary_multi_task.mp4

Half-Cheetah mixed tasks

Supplement Videos (Hopper and Walker)

walker non stationary vel.mp4
hopper_non_stationary_vel.mp4