reward error during training
显示 更早的评论
Hello,
Im using reinforcemet designer to train my model,
and here is my problem.
Q. I dont not why my reward cannot go up to 0.1, why is this happen?? How can I fix this??
回答（1 个）
Harsh
2025-2-28
0 个投票
Hi @기범
In Reinforcement Learning (RL), the reward is a signal that guides the agent’s learning by providing feedback on its actions. It changes dynamically based on the agent’s behavior and the environment’s response. A well-designed reward function encourages desired actions and discourages unwanted ones, leading to improved performance over time.
You can try the following to improve your performance:
- Ensure that “cos(psi(t)) - cos(psi(t-1)) > 0” is met frequently by checking if psi(t) increases over time.
- Verify that the initial conditions of the delay block are properly set to prevent incorrect first-step evaluations.
- Modify the reward function threshold to make the positive reward condition more lenient.
For more information on how to craft a reward function please refer to the following MATLAB tech talks webinar - https://www.mathworks.com/videos/reinforcement-learning-part-2-understanding-the-environment-and-rewards-1551976590603.html
类别
在 帮助中心 和 File Exchange 中查找有关 Environments 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
