Minor changes for better example running experience. Reset motor RPM …

…at reset. Set damping coeff for position control.

Signed-off-by: Mihir Kulkarni <[email protected]>

aerial_gym/config/robot_config/base_quad_config.py

@@ -168,22 +168,22 @@ class control_allocator_config:
         ]
 
         class motor_model_config:
-            use_rps = True
+            use_rps = False
 
             motor_thrust_constant_min = 0.00000926312
             motor_thrust_constant_max = 0.00001826312
 
-            motor_time_constant_increasing_min = 0.09
-            motor_time_constant_increasing_max = 0.12
+            motor_time_constant_increasing_min = 0.03
+            motor_time_constant_increasing_max = 0.03
 
-            motor_time_constant_decreasing_min = 0.03
-            motor_time_constant_decreasing_max = 0.05
+            motor_time_constant_decreasing_min = 0.04
+            motor_time_constant_decreasing_max = 0.04
 
             max_thrust = 2
             min_thrust = 0
 
             max_thrust_rate = 100000.0
             thrust_to_torque_ratio = 0.01
             use_discrete_approximation = (
-                False  # Setting to false will compute f' based on difference and time constant
+                True  # Setting to false will compute f' based on difference and time constant
             )

aerial_gym/control/motor_model.py

@@ -10,8 +10,12 @@ def __init__(self, num_envs, motors_per_robot, dt, config, device="cuda:0"):
         self.cfg = config
         self.device = device
         self.num_motors_per_robot = motors_per_robot
-        self.max_thrust = self.cfg.max_thrust
-        self.min_thrust = self.cfg.min_thrust
+        self.max_thrust = torch.tensor(self.cfg.max_thrust, device=self.device, dtype=torch.float32).expand(
+            self.num_envs, self.num_motors_per_robot
+        )
+        self.min_thrust = torch.tensor(self.cfg.min_thrust, device=self.device, dtype=torch.float32).expand(
+            self.num_envs, self.num_motors_per_robot
+        )
         self.motor_time_constant_increasing_min = torch.tensor(
             self.cfg.motor_time_constant_increasing_min, device=self.device
         ).expand(self.num_envs, self.num_motors_per_robot)
@@ -118,9 +122,6 @@ def reset_idx(self, env_ids):
             self.motor_thrust_constant[env_ids] = torch_rand_float_tensor(
                 self.motor_thrust_constant_min, self.motor_thrust_constant_max
             )[env_ids]
-        self.first_order_linear_mixing_factor[env_ids] = self.mixing_factor_function(
-            self.dt, self.motor_time_constants
-        )[env_ids]
 
     def reset(self):
         self.reset_idx(torch.arange(self.num_envs, device=self.device))

aerial_gym/robots/base_multirotor.py

@@ -199,6 +199,7 @@ def reset_idx(self, env_ids):
         self.robot_state[env_ids, 10:13] = random_state[env_ids, 10:13]
 
         self.controller.randomize_params(env_ids=env_ids)
+        self.control_allocator.reset_idx(env_ids)
 
         # update the states after resetting because the RL agent gets the first state after reset
         self.update_states()

aerial_gym/robots/robot_manager.py

@@ -436,6 +436,9 @@ def add_robot_to_env(
                         props["stiffness"][j_index] = self.cfg.reconfiguration_config.stiffness[
                             j_index
                         ]
+                        props["damping"][j_index] = self.cfg.reconfiguration_config.damping[
+                            j_index
+                        ]
                 elif self.cfg.reconfiguration_config.dof_mode == "velocity":
                     props["driveMode"].fill(gymapi.DOF_MODE_VEL)
                     for j_index in range(len(props["damping"])):