[Bug Report] EMAJointPositionToLimitsAction bug

[Msornerrrr]

Describe the bug

For the EMAJointPositionToLimitsAction controller, during reset(), it's trying to reset history to current joint positions. The indexing part

self._asset.data.joint_pos[env_ids, self._joint_ids]

has error if EMA control is applied for just part of the robot (e.g. arm + hand setup)

Steps to reproduce

I don't find any current robot that's using the arm + hand setup so reproducing the error would be hard. But I do have a one line fix for this issue which I will describe latter.

Fixing the error

self._prev_applied_actions[env_ids, :] = self._asset.data.joint_pos[env_ids][:, self._joint_ids]

System Info

Describe the characteristic of your environment:

• Commit: 2804ff3
• Isaac Sim Version: 2022.2.0
• OS: Ubuntu 22.04
• GPU: NVIDIA GeForce RTX 4090
• CUDA: 12.4
• GPU Driver: 550.107.02

[Mayankm96]

This seems like a broadcasting issue. A better fix probably is:

self._prev_applied_actions[env_ids, :] = self._asset.data.joint_pos[env_ids[:, None], , self._joint_ids]

Can you try if that works? If yes, can you make an MR? Thanks!

[bearpaw]

I got the same error with the following setting:

Describe the bug

In some cases, we want to move certain joints instead of all the joints when we train an agent. This is enable by passing argument env.actions.joint_pos.joint_names. This will cause an indexing error in pytorch.

Steps to reproduce

For example, we want to only move the index finger of the Allegro hand for the cube reposing task:

python source/standalone/workflows/rsl_rl/train.py  \
 --task Isaac-Repose-Cube-Allegro-v0   \
--num_env 12 \
env.actions.joint_pos.joint_names='["index_joint.*"]'

We will get the following error:

Error executing job with overrides: ['env.actions.joint_pos.joint_names=["index_joint.*"]']                                                                                                                   
Traceback (most recent call last):                                                                                                                                                                            
  File "/home/weiy/Projects/code/IsaacLab/source/extensions/omni.isaac.lab_tasks/omni/isaac/lab_tasks/utils/hydra.py", line 91, in hydra_main
    func(env_cfg, agent_cfg, *args, **kwargs)                                                                                                                                                                 
  File "/home/weiy/Projects/code/IsaacLab/source/standalone/workflows/rsl_rl/train.py", line 119, in main                    
    env = RslRlVecEnvWrapper(env)                                                                                                                                                                             
  File "/home/weiy/Projects/code/IsaacLab/source/extensions/omni.isaac.lab_tasks/omni/isaac/lab_tasks/utils/wrappers/rsl_rl/vecenv_wrapper.py", line 89, in __init__
    self.env.reset()                                                                                                                                                                                          
  File "/home/weiy/miniconda3/envs/isaaclab/lib/python3.10/site-packages/gymnasium/wrappers/order_enforcing.py", line 61, in reset
    return self.env.reset(**kwargs)                                                                                                                                                                   
  File "/home/weiy/Projects/code/IsaacLab/source/extensions/omni.isaac.lab/omni/isaac/lab/envs/manager_based_env.py", line 252, in reset                                                                      
    self._reset_idx(indices)                   
  File "/home/weiy/Projects/code/IsaacLab/source/extensions/omni.isaac.lab/omni/isaac/lab/envs/manager_based_rl_env.py", line 339, in _reset_idx                                                              
    info = self.action_manager.reset(env_ids)  
  File "/home/weiy/Projects/code/IsaacLab/source/extensions/omni.isaac.lab/omni/isaac/lab/managers/action_manager.py", line 285, in reset                                                                     
    term.reset(env_ids=env_ids)                    
  File "/home/weiy/Projects/code/IsaacLab/source/extensions/omni.isaac.lab/omni/isaac/lab/envs/mdp/actions/joint_actions_to_limits.py", line 188, in reset                                                    
    self._prev_applied_actions[env_ids, :] = self._asset.data.joint_pos[env_ids, self._joint_ids]      
IndexError: shape mismatch: indexing tensors could not be broadcast together with shapes [12], [4] 

System Info

Describe the characteristic of your environment:

• Commit: 225a896
• Isaac Sim Version: 4.2
• OS: Ubuntu 24.04.1
• GPU: RTX 3090
• CUDA: 12.4
• GPU Driver: 550.54.15

Additional context

NA

Checklist

• I have checked that there is no similar issue in the repo (required)
• I have checked that the issue is not in running Isaac Sim itself and is related to the repo

Acceptance Criteria

Add the criteria for which this task is considered done. If not known at issue creation time, you can add this once the issue is assigned.

• Criteria 1
• Criteria 2

[benrichardson28]

This bug seems to have resurfaced in Isaac Sim Version: 4.5 (now on line 206).
206: self._prev_applied_actions[env_ids, :] = self._asset.data.joint_pos[env_ids, self._joint_ids]

Steps to reproduce

If I define EMAJointPositionToLimitsAction joints=[".*"] it works fine (self._joint_ids = slice(None,None,None)), but if the joints are explicitly set, then self._joint_ids = list[int], causing self._asset.data.joint_pos[env_ids, self._joint_ids] to have two dimensions instead of three.

Fixing the error

self._prev_applied_actions[env_ids, :] = self._asset.data.joint_pos[env_ids][:,:,self._joint_ids]
or
self._prev_applied_actions[env_ids, :] = self._asset.data.joint_pos[env_ids][...,self._joint_ids]

These seem to work for both cases at least.

System Info

Describe the characteristic of your environment:

• Isaac Sim Version: 4.5
• OS: Ubuntu 22.04
• GPU: NVIDIA NVIDIA RTX 6000 Ada
• CUDA: 12.8
• GPU Driver: 570.169

[Yasu31]

I submitted a bug report about the newly resurfaced error here: #3753
and a PR that fixes it here: #3754