Custom Contact Force Observations MDP Not Working

[H-Hisamichi]

Hi everyone,

I'm working on adding foot contact force data to the observation space for legged robot reinforcement learning.
I added Code No. 1 to velocity_env_cfg.py and Code No. 2 to observations.py.

Code No. 1:

contact_forces = ObsTerm(
            func=mdp.contact_forces,
            params={"sensor_cfg": SceneEntityCfg("contact_forces", body_names="Leg.*_Toes"), "threshold": 1.0},
            noise=Unoise(n_min=-0.01, n_max=0.01),
        )

Code No. 2:

def contact_forces(env: ManagerBasedEnv, threshold: float, sensor_cfg: SceneEntityCfg) -> torch.Tensor:
    sensor: ContactSensor = env.scene.sensors[sensor_cfg.name]
    net_contact_forces = sensor.data.net_forces_w
    is_contact = torch.norm(net_contact_forces[:, :, sensor_cfg.body_ids], dim=-1)
    return is_contact

However, when I ran the RL training, I encountered Error No. 1. I tried adding print() statements for debugging, but nothing was printed, which makes me think that the custom contact_forces part of the original MDP isn't being executed at all.

Error No. 1

Error executing job with overrides: []
Traceback (most recent call last):
  File "/home/robot/IsaacLab_v2.0.2/IsaacLab/source/isaaclab/isaaclab/managers/observation_manager.py", line 92, in __init__
    self._group_obs_dim[group_name] = tuple(torch.sum(torch.stack(term_dims, dim=0), dim=0).tolist())
RuntimeError: stack expects each tensor to be equal size, but got [1] at entry 0 and [0] at entry 8

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/robot/IsaacLab_v2.0.2/IsaacLab/source/isaaclab_tasks/isaaclab_tasks/utils/hydra.py", line 101, in hydra_main
    func(env_cfg, agent_cfg, *args, **kwargs)
  File "/home/robot/IsaacLab_v2.0.2/IsaacLab/scripts/reinforcement_learning/rsl_rl/train.py", line 134, in main
    env = gym.make(args_cli.task, cfg=env_cfg, render_mode="rgb_array" if args_cli.video else None)
  File "/home/robot/isaacsim/exts/omni.isaac.ml_archive/pip_prebundle/gymnasium/envs/registration.py", line 802, in make
    env = env_creator(**env_spec_kwargs)
  File "/home/robot/IsaacLab_v2.0.2/IsaacLab/source/isaaclab/isaaclab/envs/manager_based_rl_env.py", line 79, in __init__
    super().__init__(cfg=cfg)
  File "/home/robot/IsaacLab_v2.0.2/IsaacLab/source/isaaclab/isaaclab/envs/manager_based_env.py", line 163, in __init__
    self.load_managers()
  File "/home/robot/IsaacLab_v2.0.2/IsaacLab/source/isaaclab/isaaclab/envs/manager_based_rl_env.py", line 117, in load_managers
    super().load_managers()
  File "/home/robot/IsaacLab_v2.0.2/IsaacLab/source/isaaclab/isaaclab/envs/manager_based_env.py", line 243, in load_managers
    self.observation_manager = ObservationManager(self.cfg.observations, self)
  File "/home/robot/IsaacLab_v2.0.2/IsaacLab/source/isaaclab/isaaclab/managers/observation_manager.py", line 94, in __init__
    raise RuntimeError(
RuntimeError: Unable to concatenate observation terms in group 'policy'. The shapes of the terms are: [(3,), (3,), (3,), (3,), (12,), (12,), (12,), (36,), ()]. Please ensure that the shapes are compatible for concatenation. Otherwise, set 'concatenate_terms' to False in the group configuration.

I'm not sure what I'm missing here.
Any advice would be greatly appreciated!

Thank you.

[RandomOakForest]

Thanks for starting this discussion. We'd need a longer error log to track what's causing this. Implementing contact sensor has been described here. Could you also include also links to the scripts you started from?