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

baseline-PP #63

Closed hzxBuaa closed 1 year ago

hzxBuaa commented 1 year ago

Hello, it is an excellent repo for navigation. I am wondering in your paper( https://arxiv.org/abs/2103.07119) you mentioned baseline algorithm- PP. I have two questions.

  1. The path planning part uses the Dijkstra algorithm, but what control algorithm is used by the control layer to make the vehicle track the path?
  2. Is the travel time mentioned in the paper the time it takes for the vehicle to track the path planned by Dijkstra algorithm？ I hope to receive your reply. Thank you very much.
reiniscimurs commented 1 year ago
  1. I do not recall all the details but path planner is a combination of move_base (http://wiki.ros.org/move_base) and ROSARIA (http://wiki.ros.org/ROSARIA).
  2. The time is the total execution time from the start of the exploration to reaching the goal. It is not just the tracked path as, since you do not have a prior map, it is not possible to create a proper plan before exploration. So also re-planning comes into effect.
hzxBuaa commented 1 year ago

But the Dijkstra algorithm needs a proper plan before exploring, so is it necessary to obtain all map information to adopt this algorithm?

reiniscimurs commented 1 year ago

That's not entirely so. You can obtain an initial naive plan without exploration. A grid map has known free, known occupied, and unknown elements. By using move_base we can allow planning through unknown parts of the map, so an initial plan will be obtained. Then when executing it, the unknown parts will become known, and the original plan will be re-planned. Since that takes time, we used heuristics instead and show that through the time that it takes to get to the final goal.

hzxBuaa commented 1 year ago

Thank you very much for your answer. I think I should understand what you mean. You are using the Dijkstra planning algorithm, and the control layer comes with move_base and ROSARIA. Without the need to build a map in advance, when the car is driving, the lidar senses the surrounding environment, the unknown becomes known, and then the algorithm adjusts the planned path according to the established environment, is this true?

reiniscimurs commented 1 year ago

Yes, that is how it was implemented.