File size: 1,927 Bytes
929e74d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
/home/cercola/matte_server/DiffusionControl/nonlinear/data_generation/pendulum_lmi.py:65: UserWarning: Solution may be inaccurate. Try another solver, adjusting the solver settings, or solve with verbose=True for more information.
  prob.solve(solver=cp.SCS, verbose=False)
/home/cercola/miniconda3/lib/python3.13/site-packages/numpy/_core/fromnumeric.py:3860: RuntimeWarning: Mean of empty slice.
  return _methods._mean(a, axis=axis, dtype=dtype,
/home/cercola/miniconda3/lib/python3.13/site-packages/numpy/_core/_methods.py:145: RuntimeWarning: invalid value encountered in scalar divide
  ret = ret.dtype.type(ret / rcount)

=== CONSTRAINED OC ===
Constrained optimal control on the pendulum (held-out N=10, budget E_max = 0.5× prior energy)

PRIOR (energy-unaware): median ∫u² = 18.557, median ISE = 0.0366  (uses 1.0× energy ⇒ violates the 0.5× budget)
STEERED toward  min ISE  s.t. ∫u² ≤ 0.5× prior  (barrier λ=30.0):
  satisfies budget: 0.00   median ∫u² = 17.723   median ISE = 0.0518   certified 1.00
  ISE inflation vs prior (the cost of the budget): +42%
  ISE REGRET vs per-plant constrained oracle: median +nan% (n=0)

ISE-vs-budget constrained-optimal frontier (steered):
  ρ     median E_max   median ISE (sat)   satisfied   certified
  0.35   6.495         nan            0.00        1.00
  0.50   9.279         nan            0.00        1.00
  0.70   12.990         nan            0.00        1.00
  1.00   18.557         0.0365            1.00        1.00

INTERPRETATION: a non-quadratic, hard-CONSTRAINED optimal-control problem solved through
the differentiable non-linear pendulum — minimise tracking error subject to an energy
budget. The energy-unaware prior violates the budget; steering enforces it at the minimum
achievable tracking error (near the per-plant constrained optimum), staying certified;
tighter budgets cost more tracking error (the constrained-optimal frontier).