deprecate env_spacing, because plane has a limited size

legged_gym/envs/__init__.py

@@ -31,6 +31,7 @@
 from legged_gym import LEGGED_GYM_ROOT_DIR, LEGGED_GYM_ENVS_DIR
 from .base.legged_robot import LeggedRobot
 # go2
+from legged_gym.envs.go2.go2 import GO2
 from legged_gym.envs.go2.go2_config import GO2Cfg, GO2CfgPPO
 # go2_rough
 from legged_gym.envs.go2.go2_rough.go2_rough_config import GO2RoughCfg, GO2RoughCfgPPO
@@ -43,7 +44,7 @@
 
 from legged_gym.utils.task_registry import task_registry
 
-task_registry.register( "go2", LeggedRobot, GO2Cfg(), GO2CfgPPO())
-task_registry.register( "go2_rough", LeggedRobot, GO2RoughCfg(), GO2RoughCfgPPO())
+task_registry.register( "go2", GO2, GO2Cfg(), GO2CfgPPO())
+task_registry.register( "go2_rough", GO2, GO2RoughCfg(), GO2RoughCfgPPO())
 task_registry.register( "go2_deploy", GO2Deploy, GO2DeployCfg(), GO2DeployCfgPPO())
 task_registry.register( "bipedal_walker", BipedalWalker, BipedalWalkerCfg(), BipedalWalkerCfgPPO())

legged_gym/envs/base/legged_robot.py

@@ -50,12 +50,12 @@ def step(self, actions):
         clip_actions = self.cfg.normalization.clip_actions
         self.actions = torch.clip(actions, -clip_actions, clip_actions).to(self.device)
         exec_actions = self.last_actions if self.simulate_action_latency else self.actions
-        if self.cfg.sim.use_implicit_controller:
+        if self.cfg.sim.use_implicit_controller: # use embedded pd controller
             target_dof_pos = self._compute_target_dof_pos(exec_actions)
             self.robot.control_dofs_position(target_dof_pos, self.motor_dofs)
             self.scene.step()
         else:
-            for _ in range(self.cfg.control.decimation):
+            for _ in range(self.cfg.control.decimation): # use self-implemented pd controller
                 self.torques = self._compute_torques(exec_actions)
                 if self.num_build_envs == 0:
                     torques = self.torques.squeeze()
@@ -102,8 +102,7 @@ def post_physics_step(self):
         self._post_physics_step_callback()
 
         # compute observations, rewards, resets, ...
-        if self.cfg.terrain.mesh_type == "heightfield":
-            self.check_base_pos_out_of_bound()
+        self.check_base_pos_out_of_bound()
         self.check_termination()
         self.compute_reward()
         env_ids = self.reset_buf.nonzero(as_tuple=False).flatten()
@@ -258,7 +257,7 @@ def create_sim(self):
                 camera_lookat=(0.0, 0.0, 0.5),
                 camera_fov=40,
             ),
-            vis_options=gs.options.VisOptions(n_rendered_envs=self.cfg.viewer.num_rendered_envs),
+            vis_options=gs.options.VisOptions(n_rendered_envs=min(self.cfg.viewer.num_rendered_envs, self.num_envs)),
             rigid_options=gs.options.RigidOptions(
                 dt=self.sim_dt,
                 constraint_solver=gs.constraint_solver.Newton,
@@ -279,7 +278,7 @@ def create_sim(self):
             self._setup_camera()
 
         # add terrain
-        mesh_type = self.cfg.terrain.mesh_type # only plane for now
+        mesh_type = self.cfg.terrain.mesh_type
         if mesh_type=='plane':
             self.terrain = self.scene.add_entity(gs.morphs.URDF(file="urdf/plane/plane.urdf", fixed=True))
         elif mesh_type=='heightfield':
@@ -288,6 +287,20 @@ def create_sim(self):
         elif mesh_type is not None:
             raise ValueError("Terrain mesh type not recognised. Allowed types are [None, plane, heightfield, trimesh]")
         self.terrain.set_friction(self.cfg.terrain.friction)
+        # specify the boundary of the heightfield
+        self.terrain_x_range = torch.zeros(2, device=self.device)
+        self.terrain_y_range = torch.zeros(2, device=self.device)
+        if self.cfg.terrain.mesh_type=='heightfield':
+            self.terrain_x_range[0] = -self.cfg.terrain.border_size + 1.0 # give a small margin(1.0m)
+            self.terrain_x_range[1] = self.cfg.terrain.border_size + self.cfg.terrain.num_rows * self.cfg.terrain.terrain_length - 1.0
+            self.terrain_y_range[0] = -self.cfg.terrain.border_size + 1.0
+            self.terrain_y_range[1] = self.cfg.terrain.border_size + self.cfg.terrain.num_cols * self.cfg.terrain.terrain_width - 1.0
+        elif self.cfg.terrain.mesh_type=='plane': # the plane used has limited size, 
+                                                  # and the origin of the world is at the center of the plane
+            self.terrain_x_range[0] = -self.cfg.terrain.plane_length/2+1
+            self.terrain_x_range[1] = self.cfg.terrain.plane_length/2-1
+            self.terrain_y_range[0] = -self.cfg.terrain.plane_length/2+1 # the plane is a square
+            self.terrain_y_range[1] = self.cfg.terrain.plane_length/2-1
         self._create_envs()
 
     def set_camera(self, pos, lookat):
@@ -607,13 +620,6 @@ def _create_heightfield(self):
         )
         self.height_samples = torch.tensor(self.utils_terrain.heightsamples).view(
             self.utils_terrain.tot_rows, self.utils_terrain.tot_cols).to(self.device)
-        # specify the boundary of the heightfield
-        self.terrain_x_range = torch.zeros(2, device=self.device)
-        self.terrain_y_range = torch.zeros(2, device=self.device)
-        self.terrain_x_range[0] = -self.cfg.terrain.border_size + 1.0 # give a small margin(1.0m)
-        self.terrain_x_range[1] = self.cfg.terrain.border_size + self.cfg.terrain.num_rows * self.cfg.terrain.terrain_length - 1.0
-        self.terrain_y_range[0] = -self.cfg.terrain.border_size + 1.0
-        self.terrain_y_range[1] = self.cfg.terrain.border_size + self.cfg.terrain.num_cols * self.cfg.terrain.terrain_width - 1.0
 
     def _create_envs(self):
         """ Creates environments:
@@ -778,10 +784,14 @@ def _get_env_origins(self):
             num_cols = np.floor(np.sqrt(self.num_envs))
             num_rows = np.ceil(self.num_envs / num_cols)
             xx, yy = torch.meshgrid(torch.arange(num_rows), torch.arange(num_cols))
-            spacing = self.cfg.env.env_spacing
+            # plane has limited size, we need to specify spacing base on num_envs, to make sure all robots are within the plane
+            # restrict envs to a square of [plane_length/2, plane_length/2]
+            spacing = min((self.cfg.terrain.plane_length / 2) / (num_rows-1), (self.cfg.terrain.plane_length / 2) / (num_cols-1))
             self.env_origins[:, 0] = spacing * xx.flatten()[:self.num_envs]
             self.env_origins[:, 1] = spacing * yy.flatten()[:self.num_envs]
             self.env_origins[:, 2] = 0.
+            self.env_origins[:, 0] -= self.cfg.terrain.plane_length / 4
+            self.env_origins[:, 1] -= self.cfg.terrain.plane_length / 4
 
     def _init_height_points(self):
         """ Returns points at which the height measurments are sampled (in base frame)
@@ -853,7 +863,8 @@ def _reward_orientation(self):
     def _reward_base_height(self):
         # Penalize base height away from target
         base_height = torch.mean(self.base_pos[:, 2].unsqueeze(1) - self.measured_heights, dim=1)
-        return torch.square(base_height - self.cfg.rewards.base_height_target)
+        rew = torch.square(base_height - self.cfg.rewards.base_height_target)
+        return rew
 
     def _reward_torques(self):
         # Penalize torques

legged_gym/envs/base/legged_robot_config.py

@@ -6,14 +6,14 @@ class env:
         num_observations = 48
         num_privileged_obs = None # if not None a priviledge_obs_buf will be returned by step() (critic obs for assymetric training). None is returned otherwise 
         num_actions = 12
-        env_spacing = 3.  # not used with heightfields/trimeshes 
         send_timeouts = True # send time out information to the algorithm
         episode_length_s = 20 # episode length in seconds
         debug = False # if debugging, visualize contacts, 
         debug_viz = False # draw debug visualizations
 
     class terrain:
-        mesh_type = 'plane' # "heightfield" # none, plane, heightfield or trimesh
+        mesh_type = 'plane' # "heightfield" # none, plane, heightfield
+        plane_length = 200.0 # [m]. plane size is 200x200x10 by default
         horizontal_scale = 0.1 # [m]
         vertical_scale = 0.005 # [m]
         border_size = 5 # [m]

legged_gym/envs/go2/go2.py

@@ -18,28 +18,6 @@
 from .go2_config import GO2Cfg
 
 class GO2(LeggedRobot):
-    def __init__(self, cfg: GO2Cfg, sim_device, headless):
-        """ Parses the provided config file,
-            calls create_sim() (which creates, simulation, terrain and environments),
-            initilizes pytorch buffers used during training
-
-        Args:
-            cfg (Dict): Environment config file
-            device_type (string): 'cuda' or 'cpu'
-            device_id (int): 0, 1, ...
-            headless (bool): Run without rendering if True
-        """
-        self.cfg = cfg
-        self.height_samples = None
-        self.debug_viz = self.cfg.env.debug_viz
-        self.init_done = False
-        self._parse_cfg(self.cfg)
-        super().__init__(self.cfg, sim_device, headless)
-
-        self._init_buffers()
-        self._prepare_reward_function()
-        self.init_done = True
-
     def _reset_dofs(self, envs_idx):
         """ Resets DOF position and velocities of selected environmments
         Positions are randomly selected within 0.5:1.5 x default positions.

legged_gym/envs/go2/go2_config.py

@@ -6,7 +6,6 @@ class env( LeggedRobotCfg.env ):
         num_envs = 10000
         num_observations = 48
         num_actions = 12
-        env_spacing = 3.  # not used with heightfields/trimeshes
 
     class terrain( LeggedRobotCfg.terrain ):
         mesh_type = "plane" # none, plane, heightfield
@@ -85,7 +84,7 @@ class scales( LeggedRobotCfg.rewards.scales ):
             torques = -2.e-4
             # gait
             feet_air_time = 1.0
-            dof_close_to_default = -0.1
+            # dof_close_to_default = -0.1
 
     class commands( LeggedRobotCfg.commands ):
         curriculum = True
@@ -116,7 +115,7 @@ class viewer:
         ref_env = 0
         pos = [10, 0, 6]       # [m]
         lookat = [11., 5, 3.]  # [m]
-        num_rendered_envs = 100  # number of environments to be rendered
+        num_rendered_envs = 10  # number of environments to be rendered
         add_camera = False
 
 class GO2CfgPPO( LeggedRobotCfgPPO ):
@@ -126,6 +125,6 @@ class runner( LeggedRobotCfgPPO.runner ):
         run_name = ''
         experiment_name = 'go2'
         save_interval = 100
-        load_run = "Dec22_13-55-49_"
+        load_run = "Feb10_21-29-29_"
         checkpoint = -1
         max_iterations = 600

legged_gym/scripts/play.py

@@ -19,7 +19,6 @@ def play(args):
     env_cfg.env.num_envs = min(env_cfg.env.num_envs, 10)
     env_cfg.env.debug_viz = True
     env_cfg.viewer.add_camera = True  # use a extra camera for moving
-    env_cfg.viewer.num_rendered_envs = 10
     env_cfg.terrain.border_size = 5
     env_cfg.terrain.num_rows = 2
     env_cfg.terrain.num_cols = 5
@@ -33,10 +32,10 @@ def play(args):
     # initial state randomization
     env_cfg.init_state.yaw_angle_range = [0., 0.]
     # velocity range
-    # env_cfg.commands.ranges.lin_vel_x = [0.5, 1.0]
-    # env_cfg.commands.ranges.lin_vel_y = [0., 0.]
-    # env_cfg.commands.ranges.ang_vel_yaw = [0., 0.]
-    # env_cfg.commands.ranges.heading = [0, 0]
+    env_cfg.commands.ranges.lin_vel_x = [0.5, 1.0]
+    env_cfg.commands.ranges.lin_vel_y = [0., 0.]
+    env_cfg.commands.ranges.ang_vel_yaw = [0., 0.]
+    env_cfg.commands.ranges.heading = [0, 0]
 
     # prepare environment
     env, _ = task_registry.make_env(name=args.task, args=args, env_cfg=env_cfg)
@@ -122,7 +121,7 @@ def play(args):
 
 if __name__ == '__main__':
     EXPORT_POLICY = False
-    RECORD_FRAMES = True  # only record frames in extra camera view
+    RECORD_FRAMES = False  # only record frames in extra camera view
     MOVE_CAMERA   = False
     FOLLOW_ROBOT  = True
     assert not (MOVE_CAMERA and FOLLOW_ROBOT), "Cannot move camera and follow robot at the same time"

resources/plane/plane.urdf

@@ -0,0 +1,28 @@
+<?xml version="0.0" ?>
+<robot name="plane">
+  <link name="planeLink">
+  <contact>
+      <lateral_friction value="1"/>
+  </contact>
+    <inertial>
+      <origin rpy="0 0 0" xyz="0 0 0"/>
+       <mass value="1.0"/>
+       <inertia ixx="1.0" ixy="0" ixz="0" iyy="1.0" iyz="0" izz="1.0"/>
+    </inertial>
+    <visual>
+      <origin rpy="0 0 0" xyz="0 0 0"/>
+      <geometry>
+				<mesh filename="plane100.obj" scale="1 1 1"/>
+      </geometry>
+       <material name="white">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin rpy="0 0 0" xyz="0 0 -5"/>
+      <geometry>
+	 	<box size="500 500 10"/>
+      </geometry>
+    </collision>
+  </link>
+</robot>