Reset selected robots

Run the Mujoco simulation using the provided policy and configuration. Args: policy: The policy used for controlling the simulation.

Apply actions, simulate, call self.post_physics_step() Args: actions (torch.Tensor): Tensor of shape (num_envs, num_actions_per_

Initializes all member classes recursively. Ignores all namse starting with '__' (buit-in methods).

Parses the provided config file, calls create_sim() (which creates, simulation, terrain and environments), initilizes pytorch

Initialize torch tensors which will contain simulation states and processed quantities

Callback called before computing terminations, rewards, and observations Default behaviour: Compute ang vel command based on target and h

Refresh the rigid body properties of the actor in the given environments, i.e. set the payload_mass to the desired values.

Refresh the rigid shape properties of the actor in the given environments, i.e. set the friction and restitution coefficients to the desi

Encourages smoothness in the robot's actions by penalizing large differences between consecutive actions. This is important for achie

Penalizes the use of high torques in the robot's joints. Encourages efficient movement by minimizing the necessary force exerted by t

Computes the reward based on the base's acceleration. Penalizes high accelerations of the robot's base, encouraging smoother motion.

Calculates the reward based on the robot's base height. Penalizes deviation from a target base height. The reward is computed based o

Penalizes collisions of the robot with the environment, specifically focusing on selected body parts. This encourages the robot to av

Calculates the reward for keeping joint positions close to default positions, with a focus on penalizing deviation in yaw and roll d

Penalizes high accelerations at the robot's degrees of freedom (DOF). This is important for ensuring smooth and stable motion, reduci

Penalizes high velocities at the degrees of freedom (DOF) of the robot. This encourages smoother and more controlled movements.

Calculates the reward for feet air time, promoting longer steps. This is achieved by checking the first contact with the ground after

Calculates reward based on the clearance of the swing leg from the ground during movement. Encourages appropriate lift of the feet du

Calculates the reward for keeping contact forces within a specified range. Penalizes high contact forces on the feet.

Calculates a reward based on the number of feet contacts aligning with the gait phase. Rewards or penalizes depending on whether the

Calculates the reward based on the distance between the feet. Penilize feet get close to each other or too far away.

Calculates the reward for minimizing foot slip. The reward is based on the contact forces and the speed of the feet. A contact thres

Calculates the reward based on the distance between the knee of the humanoid.

Rewards or penalizes the robot based on its speed relative to the commanded speed. This function checks if the robot is moving too s

Calculates the reward for maintaining a flat base orientation. It penalizes deviation from the desired base orientation using the ba

Calculates the reward based on the difference between the current joint positions and the target joint positions.

Penalizes the use of high torques in the robot's joints. Encourages efficient movement by minimizing the necessary force exerted by t

Calculates a reward for accurately tracking both linear and angular velocity commands. Penalizes deviations from specified linear and

Tracks angular velocity commands for yaw rotation. Computes a reward based on how closely the robot's angular velocity matches the co

Tracks linear velocity commands along the xy axes. Calculates a reward based on how closely the robot's linear velocity matches the

Computes a reward based on the mismatch in the robot's linear and angular velocities. Encourages the robot to maintain a stable velo

Creates simulation, terrain and evironments

Reset all robots

Reset all robots

Reset some environments. Calls self._reset_dofs(env_ids), self._reset_root_states(env_ids), and self._resample_commands(env_ids)