search

reiniscimurs / DRL-robot-navigation

Deep Reinforcement Learning for mobile robot navigation in ROS Gazebo simulator. Using Twin Delayed Deep Deterministic Policy Gradient (TD3) neural network, a robot learns to navigate to a random goal point in a simulated environment while avoiding obstacles.
MIT License
487 stars 97 forks source link

Robot goes around in circles. #66

Closed Talha771 closed 8 months ago

Talha771 commented 1 year ago

I cloned this repo and ran the training, however, there has been no noticable improvement in the reward and the robot just goes around rotating, and does not move at all. Am I missing setting some parameters or getting overly stressed way too early in the training process ?

image

It is moving a bit, but mainly just in circles, Also, I'm running this on an Titan RTX and it trained around 52 epochs in around 15 hours, I think its taking longer than it should, but I don't know how to change that.

reiniscimurs commented 1 year ago

Hi,

It could just e that you need to change the seed for the initialization of the network. See explanation here: https://github.com/reiniscimurs/DRL-robot-navigation/issues/19 Let me know if that does not work, then we can look into it deeper.

Talha771 commented 1 year ago

Thank you, will try that. Also thank you for being so responsive

Talha771 commented 1 year ago

This worked perfectly.

DanDan996 commented 12 months ago

This worked perfectly.

Hi, can you please tell me what your seed value is set at? I have set many values without good results and end up in the situation described in your question.

vamsi8106 commented 8 months ago

This worked perfectly.

Please may I know which value of seed has worked good I have tried many values but no improvement.