heleidsn / UAV_Navigation_DRL_AirSim

This is a new repo used for training UAV navigation (local path planning) policy using DRL methods.
212 stars 29 forks source link

SimpleAvoid训练卡住 #8

Closed Bruce-Si closed 1 year ago

Bruce-Si commented 1 year ago

您好,训练SimpleAvoid,没过一会无人机就卡在柱子上,无法reset。

微信截图_20230422173309

vechi0324 commented 1 year ago

我能问一下你在训练之前是怎样设置的么?我在训练一开始就会卡在柱子上面。

heleidsn commented 1 year ago

可以贴出一下命令行中的输出看看

Bruce-Si commented 1 year ago

我能问一下你在训练之前是怎样设置的么?我在训练一开始就会卡在柱子上面。

这是我瞎改的配置文件,别的没有设置了。 [options] project_name = FlappingWing env_name = SimpleAvoid dynamic_name = SimpleMultirotor

navigation_3d = False using_velocity_state = False reward_type = reward_final

;depth, lgmd, vector perception = vector

algo = SAC total_timesteps = 100000

policy_name = mlp net_arch = [64, 32, 16] activation_function = tanh ;cnn_feature_num = 25 cnn_feature_num = 5

keyboard_debug = False generate_q_map = True q_map_save_steps = 10000

use_wandb = False wandb_run_name = Maze-2D-mlp-tanh-M3 notes = no

;use_wandb = True ;wandb_run_name = Maze-2D-mlp-tanh-M3 ;notes = test

;[wandb] ;name = Maze-2D-mlp-tanh-M3 ;notes = test

[environment] max_depth_meters = 15 screen_height = 60 screen_width = 90

crash_distance = 2 accept_radius = 2

[multirotor] dt = 0.1 acc_xy_max = 2.0 v_xy_max = 5 v_xy_min = 0.5 v_z_max = 2.0 yaw_rate_max_deg = 30.0

; configs for DRL algorithms [DRL] gamma = 0.99 learning_rate = 1e-3 learning_starts = 2000 buffer_size = 50000 batch_size = 512 train_freq = 100 gradient_steps = 100 action_noise_sigma = 0.1

Bruce-Si commented 1 year ago

可以贴出一下命令行中的输出看看 这是命令行打印的信息,请帮看看。 微信截图_20230423114448

vechi0324 commented 1 year ago

你可以尝试一下修改C盘路径中airsim的settings.json,把仓库路径中的settings_simpledynamics.json内容贴过去试一下,好像这样就能把image_obs尺寸设置成100*80,那个函数就不会出问题了