Description

bwolleswinkel

opened on Jan 16, 2025

Issue body actions

I have found some unexplainable and unwanted behavior of the control.ctrb functions, which returns a 'controllability matrix' with incorrect dimensions rather than returning an error/warning that the $B$ matrix is not dimensioned properly.

MWE:

import numpy as np
import control as ct

A = np.array([[1, 1], [0, 1]])
b = np.array([1, 1])
C = ct.ctrb(A, b)  # Rather then 1) returning an error that b is not properly dimensioned, or 2) assuming that b is a columns vector, it instead returns a matrix with shape 2 x 4, which makes no sense for a controllability matrix

C_T = ct.ctrb(A, np.atleast_2d(b).T)  # Explicitly specifying that b is a column vector fixes this and C_T is now a 2 x 2 matrix

I believe the behavior is incorrect, and it took me quite some time to understand what was going wrong; any (dis)agreement would be appreciated, thanks.

Activity

1. roryyorke commented on Jan 19, 2025

   

   roryyorke

   on Jan 19, 2025

   Contributor

   More actions

   the arguments for ctrb and obsv aren't checked for size.

   In [42]: ct.ctrb(np.eye(2), 0.5)
   Out[42]: 
   array([[0.5, 0.5],
          [0.5, 0.5]])
   
   In [43]: ct.ctrb(0.3*np.eye(2), 0.5)
   Out[43]: 
   array([[0.5 , 0.15],
          [0.5 , 0.15]])
   
   In [44]: ct.ctrb(0.1*np.eye(2), 1)
   Out[44]: 
   array([[1. , 0.1],
          [1. , 0.1]])
   
   In [45]: ct.ctrb(0.1*np.eye(2), np.array([1,1]))
   Out[45]: 
   array([[1. , 1. , 0.1, 0.1],
          [1. , 1. , 0.1, 0.1]])
   
   In [46]: ct.obsv(0.1*np.eye(2), 1)
   Out[46]: 
   array([[1. , 1. ],
          [0.1, 0.1]])
   
   In [47]: ct.obsv(0.1*np.eye(2), [1, 1])
   Out[47]: 
   array([[1. , 1. ],
          [0.1, 0.1]])
   
   In [48]: ct.obsv(0.1*np.eye(2), [1, 1, 1])
   ---------------------------------------------------------------------------
   ValueError                                Traceback (most recent call last)
   Cell In[48], line 1
   ----> 1 ct.obsv(0.1*np.eye(2), [1, 1, 1])
   
   File ~/miniforge3/envs/work/lib/python3.12/site-packages/control/statefbk.py:1070, in obsv(A, C, t)
      1068 # Construct the observability matrix
      1069 obsv = np.zeros((t * p, n))
   -> 1070 obsv[:p, :] = cmat
      1072 for k in range(1, t):
      1073     obsv[k * p:(k + 1) * p, :] = np.dot(obsv[(k - 1) * p:k * p, :], amat)
   
   ValueError: could not broadcast input array from shape (1,3) into shape (1,2)
   

   The fix seems straightforward -- adapt the _ssmatrix calls and size checks from the statesp.StateSpace.__init__.

   Reacted by Bart Wolleswinkel
2. murrayrm

   closed this as completedin #1099on Jan 20, 2025

Issue actions