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

the velocity of robot #122

Open Ethan0207 opened 3 months ago

Ethan0207 commented 3 months ago

Hi, I changed the car model to a four wheel differential car. During training, the driving speed is very slow. The target point cannot be reached. Can you give some suggestions based on your experience?

https://github.com/reiniscimurs/DRL-robot-navigation/assets/138771150/9215b291-711f-4b41-93da-007d4aff2c95

3 6 1

reiniscimurs commented 3 months ago

Hi,

I have no input here. You should probably check velocities that are input into your car model and what are reasonable actions for a 4 wheel diff drive.

Ethan0207 commented 3 months ago

Hi， Thank you for your reply. The velocity of car model is normal. But I have an another question. During the training, when the car model hits the box, the box will also be pushed, and when it hits the wall, the wheel will be off the ground. Is this due to the large size of the car, resulting in incorrect collision detection? Do I need to adjust the reward function and the GOAL_REACHED_DIST threshold? Can you give me some advice? Looking forward to your reply.

https://github.com/reiniscimurs/DRL-robot-navigation/assets/138771150/7483f803-aab2-43b1-ac09-9e2d35d62681

3 7

reiniscimurs commented 3 months ago

Collision distance is set here: https://github.com/reiniscimurs/DRL-robot-navigation/blob/main/TD3/velodyne_env.py#L21

If the front of the robot is further than this distance, the collision will not be detected.