Continuous-Time RL
Paper Codebase
What you will find here. Trading qualitative plots and videos, irregular-time control images and videos, and regular-time transfer media for policies trained under irregular decision times.

Qualitative results for the trading environment, including single-policy rollouts and direct comparisons between CT-SAC and competitive baselines.

Trading qualitative figures


Trading videos

CT-SAC vs SAC

Side-by-side comparison between CT-SAC and SAC.

CT-SAC vs CPPO

Side-by-side comparison between CT-SAC and CPPO.

Irregular-time control

Qualitative control results under the same irregular-time setting used in training and evaluation. The static images represent the last outcome, while the videos show representative rollout behavior.

Irregular-time qualitative figures







Irregular-time videos

Our method against continuous-time baselines on Cheetah

Irregular-time evaluation on Cheetah against continuous-time baselines.

Our method against discrete-time baselines on Cheetah

Irregular-time evaluation on Cheetah against discrete-time baselines.

Our method against continuous-time baselines on Walker

Irregular-time evaluation on Walker against continuous-time baselines.

Our method against discrete-time baselines on Walker

Irregular-time evaluation on Walker against discrete-time baselines.

Our method against continuous-time baselines on Humanoid

Irregular-time evaluation on Humanoid against continuous-time baselines.

Our method against discrete-time baselines on Humanoid

Irregular-time evaluation on Humanoid against discrete-time baselines.

Our method against continuous-time baselines on Quadruped

Irregular-time evaluation on Quadruped against continuous-time baselines.

Our method against discrete-time baselines on Quadruped

Irregular-time evaluation on Quadruped against discrete-time baselines.

Regular-time transfer

Qualitative control test results under the regular-time setting (training on irregular, evaluation on regular). The static images represent the last outcome, while the videos show representative rollout behavior.

Regular-time qualitative figures







Regular-time videos

Our method against continuous-time baselines on Cheetah

Regular-time evaluation on Cheetah after irregular-time training, against continuous-time baselines.

Our method against discrete-time baselines on Cheetah

Regular-time evaluation on Cheetah after irregular-time training, against discrete-time baselines.

Our method against continuous-time baselines on Walker

Regular-time evaluation on Walker after irregular-time training, against continuous-time baselines.

Our method against discrete-time baselines on Walker

Regular-time evaluation on Walker after irregular-time training, against discrete-time baselines.

Our method against continuous-time baselines on Humanoid

Regular-time evaluation on Humanoid after irregular-time training, against continuous-time baselines.

Our method against discrete-time baselines on Humanoid

Regular-time evaluation on Humanoid after irregular-time training, against discrete-time baselines.

Our method against continuous-time baselines on Quadruped

Regular-time evaluation on Quadruped after irregular-time training, against continuous-time baselines.

Our method against discrete-time baselines on Quadruped

Regular-time evaluation on Quadruped after irregular-time training, against discrete-time baselines.