12345qiupeng
/
mpc_python_learn
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

add notes on how to improve controller

master
mcarfagno 2023-10-24 11:46:56 +01:00
parent 7369fedcb3
commit 6239e769b1
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
mpc_pybullet_demo/cvxpy_mpc/cvxpy_mpc.py
View File

@ -105,12 +105,16 @@ class MPC:
assert len(initial_state) == self.nx assert len(initial_state) == self.nx
# Create variables # Create variables needed for setting up cvxpy problem
x = opt.Variable((self.nx, self.control_horizon + 1), name="states") x = opt.Variable((self.nx, self.control_horizon + 1), name="states")
u = opt.Variable((self.nu, self.control_horizon), name="actions") u = opt.Variable((self.nu, self.control_horizon), name="actions")
cost = 0 cost = 0
constr = [] constr = []
# NOTE: here the state linearization is performed around the starting condition to simplify the controller.
# This approximation gets more inaccurate as the controller looks at the future.
# To improve performance we can keep track of previous optimized x, u and compute these matrices for each timestep k
# Ak, Bk, Ck = self.get_linear_model_matrices(x_prev[:,k], u_prev[:,k])
A, B, C = self.get_linear_model_matrices(initial_state, prev_cmd) A, B, C = self.get_linear_model_matrices(initial_state, prev_cmd)
for k in range(self.control_horizon): for k in range(self.control_horizon):