How to add observation term from height map sensor in the deploying policy script

Omniverse Isaac Sim
, ,
1

Isaac Sim Version

4.5.0

Operating System

Ubuntu 22.04

Topic Description

I have trained a locomotion policy with the Unitree Go2 robot, following the example in velocity_env_cfg.py, and obtained the policy.pt and env.yaml files in the exported results. Then, I tried to follow the Deploying Policies in Isaac Sim documentation and examples in standalone_examples.api.isaacsim.robot.policy.examples and the policy controller example in exts.isaacsim.robot.policy.examples.robots.

However, inside the _compute_observation() method, there are no examples of observation terms from the height scanner, which is included in my PolicyCfg in Isaac Lab. So far, I haven’t found any documentation showing this kind of example. I would like to know if it’s possible to deploy the policy in IsaacSim with height scanner observations in a simple way.

Here is the _compute_observation from anymal configuration.

  def _compute_observation(self, command):
      """
      Computes the the observation vector for the policy

      Argument:
      command (np.ndarray) -- the robot command (v_x, v_y, w_z)

      Returns:
      np.ndarray -- The observation vector.

      """
      lin_vel_I = self.robot.get_linear_velocity()
      ang_vel_I = self.robot.get_angular_velocity()
      pos_IB, q_IB = self.robot.get_world_pose()

      R_IB = quat_to_rot_matrix(q_IB)
      R_BI = R_IB.transpose()
      lin_vel_b = np.matmul(R_BI, lin_vel_I)
      ang_vel_b = np.matmul(R_BI, ang_vel_I)
      gravity_b = np.matmul(R_BI, np.array([0.0, 0.0, -1.0]))

      # height_map

      obs = np.zeros(48)
      # Base lin vel
      obs[:3] = lin_vel_b
      # Base ang vel
      obs[3:6] = ang_vel_b
      # Gravity
      obs[6:9] = gravity_b
      # Command
      obs[9:12] = command
      # Joint states
      current_joint_pos = self.robot.get_joint_positions()
      current_joint_vel = self.robot.get_joint_velocities()
      obs[12:24] = current_joint_pos - self.default_pos
      obs[24:36] = current_joint_vel

      # Previous Action
      obs[36:48] = self._previous_action

      return obs
3

Thank you for your interest in Isaac Lab. To ensure efficient support and collaboration, please submit your topic to its GitHub repo following the instructions provided on Isaac Lab’s Contributing Guidelines regarding discussions, submitting issues, feature requests, and contributing to the project.

We appreciate your understanding and look forward to assisting you.