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
486 stars 97 forks source link

Lose efficacy: removing all obstacles and navigating #112

Closed hzxBuaa closed 3 months ago

hzxBuaa commented 4 months ago

Hello, may I ask if it is normal for me to use a trained model in testing mode to find the endpoint, but after removing all obstacles and navigating on the open ground, I cannot find the endpoint?

reiniscimurs commented 4 months ago

Hi,

Not sure what the setup here is. What do you mean with not being able to find the endpoint? Is it not able to navigate after removing all obstacles?

If that is the case, my guess is that all laser readings are returning maximum value (10 by default) which is a situation not encountered in training. As such, the policy does not know how to respond to such out of distribution scenarionand policy performance degrades.

hzxBuaa commented 3 months ago

OK thank you very much for your answer