search

shenmuxin / uav_tracking_metaRL

This is a work on using meta-reinforcement learning to achieve autonomous target tracking and obstacle avoidance for UAVs
24 stars 3 forks source link

Unknown model iris_UST10LX #2

Open shenqg168 opened 7 months ago

shenqg168 commented 7 months ago

Muxin, hello. I encountered the following error while reproducing your code: I have followed the installation steps as follows: Copy the robotic model located in the /models directory of this project to the /PX4-Autopilot/Tools/sitl_gazebo/models directory. There was an issue encountered while executing roslaunch uav_tracking_metaRL setup_2uavs.launch. Error: Unknown model iris_UST10LX Can you please tell me how to solve it.Looking forward to your reply！

shenmuxin commented 7 months ago

About this issue, you can follow the (3.2 Modification in PX4) in the Readme.md.

Specifically speaking, you should add the two custom models in the path /models to the storage path of PX4, which are imu_gazebo and iris_UST10LX. After doing this, you should modify some files in PX4 and recompile, so that PX4 can find the custom models. To avoid redundancy, I've omitted the details of adding a custom model. You can follow this answer carefully.

Hope this can be helpful.

1710127727386

shenqg168 commented 7 months ago

Thank you very much! The model issue has been resolved, but a new problem has emerged. When running rosrun uav_tracking_metaRL pearl_trainer_node.py in the Conda virtual environment, the following error occurs: `***** [finish test]

Collecting initial pool of data for train and eval 0%| | 0/3 [00:00<?, ?it/s]Current task index is 0 vx:0.6 yaw_rate:-0.04

[INFO] [1710222725.328393, 291.362000]: >>>>>> Generate Cylinders >>>>>> [INFO] [1710222744.859576, 310.864000]: Begin Game Current task index is 0 vx:0.6 yaw_rate:-0.04 0%| | 0/3 [00:48<?, ?it/s] Traceback (most recent call last): File "/home/sj/shen_ws/rl_drone/catkin_ws/devel/lib/uav_tracking_metarl/pearl_trainer_node.py", line 15, in exec(compile(fh.read(), python_script, 'exec'), context) File "/home/sj/shen_ws/rl_drone/catkin_ws/src/uav_tracking_metarl/scripts/pearl2buf/pearl_trainer_node.py", line 91, in meta_learner.meta_train() File "/home/sj/shen_ws/rl_drone/catkin_ws/src/uav_tracking_metarl/scripts/pearl2buf/algorithm/meta_learner.py", line 225, in meta_train self.collect_train_data( File "/home/sj/shen_ws/rl_drone/catkin_ws/src/uav_tracking_metarl/scripts/pearl2buf/algorithm/meta_learner.py", line 194, in collect_train_data trajs, num_samples = self.sampler.obtain_samples( File "/home/sj/shen_ws/rl_drone/catkin_ws/src/uav_tracking_metarl/scripts/pearl2buf/algorithm/sampler.py", line 46, in obtain_samples traj = self.rollout(task_index=task_index, accum_context=accum_context) File "/home/sj/shen_ws/rl_drone/catkin_ws/src/uav_tracking_metarl/scripts/pearl2buf/algorithm/sampler.py", line 76, in rollout action = self.agent.get_action(obs) File "/home/sj/shen_ws/rl_drone/catkin_ws/src/uav_tracking_metarl/scripts/pearl2buf/algorithm/sac.py", line 97, in getaction action, = self.policy(inputs) File "/home/sj/anaconda3/envs/env_meta/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, *kwargs) File "/home/sj/shen_ws/rl_drone/catkin_ws/src/uav_tracking_metarl/scripts/pearl2buf/algorithm/networks.py", line 193, in forward x = self.hidden_activation(fc_layer(x)) File "/home/sj/anaconda3/envs/env_meta/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(args, **kwargs) File "/home/sj/anaconda3/envs/env_meta/lib/python3.8/site-packages/torch/nn/modules/linear.py", line 114, in forward return F.linear(input, self.weight, self.bias) RuntimeError: mat1 and mat2 shapes cannot be multiplied (1x17 and 52x300) ` Sometimes when starting PX4 and Gazebo, the following error may be prompted: ERROR [simulator] poll timeout 0, 22 ERROR [simulator] poll timeout 0, 22 ERROR [simulator] poll timeout 0, 22 ERROR [simulator] poll timeout 0, 22 ERROR [simulator] poll timeout 0, 22 Can you please tell me how to solve it? Looking forward to your reply！

shenmuxin commented 7 months ago

File "/home/sj/shen_ws/rl_drone/catkin_ws/src/uav_tracking_metarl/scripts/pearl2buf/algorithm/networks.py", line 193, in forward x = self.hidden_activation(fc_layer(x)) RuntimeError: mat1 and mat2 shapes cannot be multiplied (1x17 and 52x300)

Maybe the dimension of the nueral network is incorrect. The dimension of the state is 47, and the dimension of the action is 2. You should check it out carefully.

darlingbenke commented 7 months ago

I couldn't find the paper. Could you provide it