How to deal with failed trajectories when using record_sim_episodes.py in the simulation environment?

[return-sleep]

I've been using the record_sim_episodes.py script to generate trajectories in the simulation environment. However, I've noticed that many of the generated trajectories are failed attempts.

episode_idx=0 Failed
Replaying joint commands
Rollout out EE space scripted policy
episode_idx=1 Failed
Replaying joint commands
Rollout out EE space scripted policy
episode_idx=2 Failed
Replaying joint commands
episode_idx=2 Failed
Saving: 0.6 secs

I have two questions:
Should I filter out these failed trajectories when using them for imitation learning?
In my understanding, failed trajectories could degrade the performance of the learned policy in imitation learning. What is the recommended approach to handle this?

Thank you for your help!

[28734-mcke]

I want to konw it!!!