It seems get_low_dim_data() method has bug inside

[mch5048]

Hi, I'm really enjoying this fascinating environment you guys have released to facilitate the RL robotics research.
While I'm implementing some algorithms that is going to be tested on your environment, I encountered an error related to get_low_dim_data() method of Observation() class.
In
low_dim_data = [] if self.gripper_open is None else [self.gripper_open] for data in [self.joint_velocities, self.joint_positions, self.joint_forces, self.gripper_pose, self.gripper_joint_positions, self.gripper_touch_forces, self.task_low_dim_state]: if data is not None: low_dim_data.extend(data) return np.concatenate(low_dim_data),

the np.concatenate does not work since the gripper_open attribute is float, while others are np.ndarray, resulting the python complaining as

return np.concatenate(low_dim_data)
ValueError: zero-dimensional arrays cannot be concatenated

Is this a minor bug or am I using it in the wrong way?

Thanks in advance!

[mch5048]

low_dim_data = [] if obs.gripper_open is None else [np.array(self.gripper_open).reshape(-1)] for data in [self.joint_velocities, self.joint_positions, self.joint_forces, self.gripper_pose, self.gripper_joint_positions, self.gripper_touch_forces, self.task_low_dim_state]: if data is not None: low_dim_data.append(data) for data in low_dim_data: return np.concatenate(low_dim_data)

I changed to code as above and it seems like working fine!

[stepjam]

Thanks for pointing out. Fixed in #15.