Constantly high negative reward in RL agent
1 vue (au cours des 30 derniers jours)
Afficher commentaires plus anciens
Apoorv Pandey
le 12 Jan 2023
Commenté : awcii
le 4 Août 2023
I am using a non linear controller with an RL agent to perform hovering of a quadrotor but the reward recieved is constantly -20000 (I have given -100 as negative reward and max episode length as 200). There is no change in the z position of the UAV also please help.
Thank you in advance.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1260750/image.png)
Réponse acceptée
Emmanouil Tzorakoleftherakis
le 24 Jan 2023
You need to see what the actions generated by the RL Agent block are and how they affect the quadrotor dynamics. That's what it comes down to. Also, if you are saying that z does not change at all, that makes me think there is a modeling error as well (unless the quadrotor can hover in place by default and and output of the agent is zero)
0 commentaires
Plus de réponses (0)
Voir également
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!