Research Article
Intelligent Online Multiconstrained Reentry Guidance Based on Hindsight Experience Replay
Table 4
The hyperparameters in the training.
| | Hyperparameter | DDPG | PPO | DDPG+HER |
| | Discount factor | 0.99 | 0.99 | 0.99 | | Batch size | 64 | 64 | 64 | | Replay buffer size | 20000 | — | 20000 | | Actor learning rate | 10−4 | 10−3 | 10−4 | | Critic learning rate | 10−3 | 10−3 | 10−3 | | Target update rate | 0.001 | — | 0.001 | | Maximum number of steps | 1000 | 1000 | 1000 | | Exploration policy | OU | — | | | GAE factor | — | 0.98 | — | | Clip factor | — | 0.2 | — |
|
|
