Updating configs and cleaning up nl_mpsc

Federico-PizarroBejarano · Dec 6, 2024 · 01fc8d1 · 01fc8d1
1 parent 076393f
commit 01fc8d1
Show file tree

Hide file tree

Showing 5 changed files with 8 additions and 26 deletions.
diff --git a/examples/rl/config_overrides/quadrotor_2D_attitude/ppo_quadrotor_2D_attitude.yaml b/examples/rl/config_overrides/quadrotor_2D_attitude/ppo_quadrotor_2D_attitude.yaml
@@ -1,8 +1,8 @@
 algo: ppo
 algo_config:
   # model args
-  hidden_dim: 128
-  activation: tanh
+  hidden_dim: 64
+  activation: relu
 
   # loss args
   gamma: 0.98
@@ -19,7 +19,7 @@ algo_config:
   critic_lr: 0.001
 
   # runner args
-  max_env_steps: 2640000
+  max_env_steps: 660000
   rollout_batch_size: 1
   rollout_steps: 660
   eval_batch_size: 10

diff --git a/examples/rl/config_overrides/quadrotor_2D_attitude/quadrotor_2D_attitude_track.yaml b/examples/rl/config_overrides/quadrotor_2D_attitude/quadrotor_2D_attitude_track.yaml
@@ -34,18 +34,10 @@ task_config:
   # RL Reward
   rew_state_weight: [10, 0.1, 10, 0.1, 0.1, 0.001]
   rew_act_weight: [0.1, 0.1]
+  info_mse_metric_state_weight: [1, 0, 1, 0, 0, 0]
   rew_exponential: True
 
-  # disturbances:
-  #   observation:
-  #     - disturbance_func: white_noise
-  #       std: [5.6e-05, 1.5e-04, 2.9e-05, 8.0e-04, 1.3e-04, 3.6e-04, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
-
   constraints:
-    - constraint_form: default_constraint
-      constrained_variable: state
-      upper_bounds: [ 0.9,  2, 1.45,  2,  0.75,  3]
-      lower_bounds: [-0.9, -2, 0.55, -2, -0.75, -3]
     - constraint_form: default_constraint
       constrained_variable: input
 

diff --git a/experiments/mpsc/config_overrides/ppo_quadrotor_2D_attitude.yaml b/experiments/mpsc/config_overrides/ppo_quadrotor_2D_attitude.yaml
@@ -1,8 +1,8 @@
 algo: ppo
 algo_config:
   # model args
-  hidden_dim: 128
-  activation: tanh
+  hidden_dim: 64
+  activation: relu
 
   # loss args
   gamma: 0.98
@@ -19,7 +19,7 @@ algo_config:
   critic_lr: 0.001
 
   # runner args
-  max_env_steps: 2640000
+  max_env_steps: 660000
   rollout_batch_size: 1
   rollout_steps: 660
   eval_batch_size: 10

diff --git a/experiments/mpsc/config_overrides/quadrotor_2D_attitude_tracking.yaml b/experiments/mpsc/config_overrides/quadrotor_2D_attitude_tracking.yaml
@@ -34,13 +34,9 @@ task_config:
   # RL Reward
   rew_state_weight: [10, 0.1, 10, 0.1, 0.1, 0.001]
   rew_act_weight: [0.1, 0.1]
+  info_mse_metric_state_weight: [1, 0, 1, 0, 0, 0]
   rew_exponential: True
 
-  # disturbances:
-  #   observation:
-  #     - disturbance_func: white_noise
-  #       std: [5.6e-05, 1.5e-04, 2.9e-05, 8.0e-04, 1.3e-04, 3.6e-04, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
-
   constraints:
     - constraint_form: default_constraint
       constrained_variable: state

diff --git a/safe_control_gym/safety_filters/mpsc/nl_mpsc.py b/safe_control_gym/safety_filters/mpsc/nl_mpsc.py
@@ -990,26 +990,20 @@ def setup_acados_optimizer(self):
 
         # set cost module
         ocp.cost.cost_type = 'LINEAR_LS'
-        ocp.cost.cost_type_e = 'LINEAR_LS'
 
         Q_mat = np.zeros((nx, nx))
-        ocp.cost.W_e = np.zeros((nx, nx))
         R_mat = np.eye(nu)
         ocp.cost.W = block_diag(Q_mat, R_mat)
 
         ocp.cost.Vx = np.zeros((ny, nx))
         ocp.cost.Vu = np.zeros((ny, nu))
         ocp.cost.Vu[nx:nx + nu, :] = np.eye(nu)
-        ocp.cost.Vx_e = np.eye(nx)
 
         # Updated on each iteration
         ocp.cost.yref = np.concatenate((self.model.X_EQ, self.model.U_EQ))
-        ocp.cost.yref_e = self.model.X_EQ
 
         # set constraints
         ocp.constraints.constr_type = 'BGH'
-        ocp.constraints.constr_type_e = 'BGH'
-
         ocp.constraints.x0 = self.model.X_EQ
         ocp.constraints.C = self.L_x
         ocp.constraints.D = self.L_u