Question: the observation space of quadruped robot

[swstbecrpgmail]

In the paper, it says that ""The input to the RL agent is the heading direction, which is translated to an angular rotation, and linear velocities commands for the x- and y-axis expressed in its body frame." May I ask that do you really not include height map information into the robot's observation space when training quadruped robot for complex terrain locomotion tasks?