problems with distributed consensus ADMM example using CVXPY #2143

Jushan-Randy-Chen · 2023-06-08T10:20:00Z

Jushan-Randy-Chen
Jun 8, 2023

Hi,

I was trying to run the consensus ADMM model predictive control based on the sample code on this page. Basically, I have a multi-agent MPC problem where each agent has 6 states and 3 inputs, and a finite prediction horizon of T, and thus each objective is summed over the horizon. Ultimately, the problem I'm trying to solve is given as:

But to keep things simple to start with, I am only considering the consensus constraint as described on this page. Here are my modified code:


from multiprocessing import Process, Pipe
import cvxpy as cp
# Number of terms f_i.
N = 3
# A list of all the f_i.
T = 10 #Prediction horizon
n_states = 6
n_inputs = 3
Q_try = np.eye(6)
R_try = np.eye(3)*0.1

x = cp.Variable((T+1,nx))
u_input = cp.Variable((T,nu))
n = n_states*N

f_list = []
cost = 0
constr = []


for id in range(n_agents):
    for t in range(T):
        cost += cp.quad_form(x[t+1,id*n_states:(id+1)*n_states],Q_try) \
                + cp.quad_form(u_input[t,id*n_inputs:(id+1)*n_inputs],R_try)
    f_list.append(cost+cp.quad_form(x[-1, id*n_states:(id+1)*n_states],Q_try*100))
    
def run_worker(f, pipe):
    rho = 1
    xbar = cp.Parameter(n, value=np.zeros(n))
    u = cp.Parameter(n, value=np.zeros(n))
    for horizon in range(T):
        f += (rho/2)*cp.sum_squares(x[horizon] - xbar + u[horizon])
    prox = cp.Problem(cp.Minimize(f))
    # ADMM loop.
    iter = 0
    while iter <= T:
        iter +=1
        prox.solve(solver=cp.OSQP)
        pipe.send(x.value)
        xbar.value = pipe.recv()
        u.value[iter] += x.value[iter] - xbar.value

# Setup the workers.
pipes = []
procs = []
for i in range(N):
    local, remote = Pipe()
    pipes += [local]
    procs += [Process(target=run_worker, args=(f_list[i], remote))]
    procs[-1].start()

# ADMM loop.
MAX_ITER=50
for i in range(MAX_ITER):
    # Gather and average xi
    xbar = sum(pipe.recv() for pipe in pipes)/N
    # Scatter xbar
    for pipe in pipes:
        pipe.send(xbar)

[p.terminate() for p in procs]

However, when I run the code the following error pops up:

Process Process-31:
Process Process-32:
Process Process-33:
Traceback (most recent call last):
Traceback (most recent call last):
Traceback (most recent call last):
  File "/home/randychen233/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/process.py", line 315, in _bootstrap
    self.run()
  File "/home/randychen233/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/process.py", line 315, in _bootstrap
    self.run()
  File "/home/randychen233/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/process.py", line 315, in _bootstrap
    self.run()
  File "/home/randychen233/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/home/randychen233/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/home/randychen233/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/tmp/ipykernel_29488/1044615635.py", line 40, in run_worker
    xbar.value = pipe.recv()
  File "/tmp/ipykernel_29488/1044615635.py", line 40, in run_worker
    xbar.value = pipe.recv()
  File "/tmp/ipykernel_29488/1044615635.py", line 40, in run_worker
    xbar.value = pipe.recv()
  File "/home/randychen233/anaconda3/envs/ICON_lab/lib/python3.10/site-packages/cvxpy/expressions/constants/parameter.py", line 87, in value
...
    raise ValueError(
ValueError: Invalid dimensions (11, 18) for Parameter value.
ValueError: Invalid dimensions (11, 18) for Parameter value.
ValueError: Invalid dimensions (11, 18) for Parameter value.
Output is truncated. View as a [scrollable element](command:cellOutput.enableScrolling?c88d6d30-2bc5-408b-9d34-69e63b505436) or open in a [text editor](command:workbench.action.openLargeOutput?c88d6d30-2bc5-408b-9d34-69e63b505436). Adjust cell output [settings](command:workbench.action.openSettings?%5B%22%40tag%3AnotebookOutputLayout%22%5D)...
---------------------------------------------------------------------------
EOFError                                  Traceback (most recent call last)
/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb Cell 33 in ()
     [53](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=52) MAX_ITER=50
     [54](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=53) for i in range(MAX_ITER):
     [55](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=54)     # Gather and average xi
---> [56](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=55)     xbar = sum(pipe.recv() for pipe in pipes)/N
     [57](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=56)     # Scatter xbar
     [58](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=57)     for pipe in pipes:

/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb Cell 33 in (.0)
     [53](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=52) MAX_ITER=50
     [54](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=53) for i in range(MAX_ITER):
     [55](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=54)     # Gather and average xi
---> [56](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=55)     xbar = sum(pipe.recv() for pipe in pipes)/N
     [57](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=56)     # Scatter xbar
     [58](vscode-notebook-cell://wsl%2Bubuntu-20.04/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/scratch.ipynb#Z1233sdnNjb2RlLXJlbW90ZQ%3D%3D?line=57)     for pipe in pipes:

File [~/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/connection.py:255](https://vscode-remote+wsl-002bubuntu-002d20-002e04.vscode-resource.vscode-cdn.net/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/~/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/connection.py:255), in _ConnectionBase.recv(self)
    253 self._check_closed()
    254 self._check_readable()
--> 255 buf = self._recv_bytes()
    256 return _ForkingPickler.loads(buf.getbuffer())

File [~/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/connection.py:419](https://vscode-remote+wsl-002bubuntu-002d20-002e04.vscode-resource.vscode-cdn.net/home/randychen233/projects/ICON_lab/distributed_opt/distributed_ADMM/~/anaconda3/envs/ICON_lab/lib/python3.10/multiprocessing/connection.py:419), in Connection._recv_bytes(self, maxsize)
...
--> 388         raise EOFError
    389     else:
    390         raise OSError("got end of file during message")

EOFError:

I don't know what exactly is causing the issue. Since my state and control vector over the horizon T are 2-dimensional, should I use a 2-dimensional parameter variable to fix this issue?
Any help is greatly appreciatd!

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

problems with distributed consensus ADMM example using CVXPY #2143

Uh oh!

{{title}}

Uh oh!

Replies: 0 comments

Select a reply

Uh oh!

problems with distributed consensus ADMM example using CVXPY #2143

Uh oh!

Jushan-Randy-Chen Jun 8, 2023

Replies: 0 comments

Jushan-Randy-Chen
Jun 8, 2023