How many iterations do we need during the training phase? I ended this 1500 times. But the test results showed that he couldn't reach the target point.

zerosansan / td3_ddpg_sac_dqn_qlearning_sarsa_mobile_robot_navigation

Reinforcement Learning (RL) based navigation implementation for mobile robot navigation. The algorithms of TD3, DDPG, SAC, DQN, Q-Learning, and SARSA were implemented.

18 stars 1 forks source link

你好。根据我的实验，机器人应该在达到第 300 集之前就开始收敛。

https://github.com/ailabspace/drl-based-mapless-crowd-navigation-with-perceived-risk/blob/main/turtlebot3_rl_sim/src/results/td3/revamped/new_tracking_cp_gcp_nobonus_corrected_3/td3_training.csv

这是我在训练期间获得的原始日志文件。

Can it be 100% successful in the end? I trained it but it didn't succeed

Hi. Based on my experiments, the robot should start converging before it reaches episode 300.

https://github.com/ailabspace/drl-based-mapless-crowd-navigation-with-perceived-risk/blob/main/turtlebot3_rl_sim/src/results/td3/revamped/new_tracking_cp_gcp_nobonus_corrected_3/td3_training.csv

Here are the original log files that I got during training.

Can it be 100% successful in the end? I trained it but it didn't succeed

zerosansan / td3_ddpg_sac_dqn_qlearning_sarsa_mobile_robot_navigation

How many iterations do we need during the training phase? I ended this 1500 times. But the test results showed that he couldn't reach the target point. #5