search

mudassirej / Vision-Based-Autonomous-Navigation-Approach-for-a-Tracked-Robot-Using-Deep-Reinforcement-Learning

4 stars 0 forks source link

the trained models are transferable from simulation to reality #1

Open an180 opened 2 years ago

an180 commented 2 years ago

Hi, I am a beginner in reinforcement learning. Thank you for sharing your paper, it has helped me a lot. I recently tried to port a model trained in gazebo to a real mobile bot. The robot has no problem subscribing and posting topics, but the robot is spinning in place, hesitating and unable to move forward. I wonder if this is caused by the difference between the simulation and the real environment, so I would like to ask you if you have encountered similar problems or if you know how to solve this?

mudassirej commented 1 year ago

hi. Yes. because real images and quite different from simulation images. you have to make sure that the camera config is the same and also you have to check the actions against the image. Do one thing, check the action classes against the image, and then you will see what action a robot performs for a similar image in the gazebo and the real world. Please let me know if you need further help. Sorry for the late reply