Home / Docs-Technical WhitePaper / 25-EFT.WP.STG.Dynamics v1.0
Chapter 8 — Controllability & Observability (Structural / Energy-Based)
One-Sentence Goal
On G = (V, E), establish structural and energy-based notions of controllability/observability, and deliver computable workflows from graph priors to Gramians and PBH tests—with dual-form checks, contracts, and runtime panel fields.
I. Scope and Objects
- Objects
- Continuous/discrete network dynamics: dx/dt = A x + B u, y = C x + n; or x_{k+1} = A_d x_k + B_d u_k, y_k = C x_k + n_k.
- Graph priors: A may be derived from −L, A_sym, A_norm, or from identified models (Ch. 7); B’s driver set and C’s sensor set are selected by placement policies.
- Inputs
G, A/L, B?, C?, sampling Δt, window [t_0, t_1], RefCond, and unit system. - Outputs
Structural controllability/observability decisions and minimal driver/sensor suggestions; energy metrics (Gramians, average/modal controllability); delta_form_ctrl/obs; manifest.stg.ctrlobs. - Constraints
- unit(x), unit(u), unit(y); unit(dx/dt) = unit(x)/[T]; check_dim must pass.
- Stability assumption: infinite-horizon Gramians require A Hurwitz (or ρ(A_d) < 1).
II. Terms and Variables
- Graph & operators: incidence B_inc, Laplacian L, system matrix A, discrete A_d = expm(A Δt).
- Sets & selection: driver nodes U ⊆ V, sensor nodes S ⊆ V; selection matrices B = S_U, C = S_S^T.
- Gramians: W_c(T) = ( ∫_0^T e^{A τ} B B^T e^{A^T τ} dτ ), W_o(T) = ( ∫_0^T e^{A^T τ} C^T C e^{A τ} dτ ).
- Structural quantities: maximum matching M*, unmatched nodes U* (driver lower bound); strongly connected components (SCCs).
- Indicators: average controllability AC, modal controllability MC, condition number κ(W), minimal energy E_min.
III. Axioms P708-*
- P708-1 (Parallel dual forms) — Evaluate structural (matching/SCC) and energy (Gramian/PBH) criteria in parallel and record discrepancies.
- P708-2 (PBH baseline) — Controllability via PBH: for all λ ∈ spec(A), rank([ λI − A, B ]) = N; observability is the dual test.
- P708-3 (Explicit measures/domains) — For Gramians, write ( ∫_0^T • dτ ); for discrete, ( Σ_{k=0}^{K−1} • ).
- P708-4 (Units/dimensions) — unit(W_c) = unit(x)^2 • [T]; unit(E_min) = unit(u)^2 • [T]; enforce check_dim.
- P708-5 (Topological consistency) — If using L / A from Chs. 2/4/7, persist construction method and hashes.
- P708-6 (Robustness) — Prefer regularized/truncated inverses for W^{-1}; report κ(W) and pseudo-inverse thresholds.
IV. Minimal Equations S708-*
- S708-1 (Kalman matrices):
Continuous controllability: Ctrb = [ B, A B, A^2 B, …, A^{N−1} B ], rank(Ctrb) = N ⇔ controllable.
Observability: Obsv = [ C^T, A^T C^T, …, (A^T)^{N−1} C^T ]^T, rank(Obsv) = N. - S708-2 (PBH):
Controllability: rank([ λI − A, B ]) = N, ∀ λ ∈ spec(A);
Observability: rank([ λI − A; C ]) = N. - S708-3 (Gramian & Lyapunov equivalence):
Infinite horizon, A Hurwitz: - A W_c + W_c A^T + B B^T = 0,
- A^T W_o + W_o A + C^T C = 0.
Finite horizon uses the integral definitions above.
Dual-form gap: delta_form_ctrl = || W_c(T) − W_c^{Lyap}(T) ||_F (discrete analog likewise).
- S708-4 (Minimum energy):
E_min(x_0 → x_T) = r^T W_c(T)^{−1} r, with r = x_T − e^{A T} x_0.
To the origin: E_min(x_0 → 0) = x_0^T W_c(T)^{−1} x_0. - S708-5 (Average / modal controllability):
Average: AC = trace(W_c(T)) or single-input AC_i = trace(W_c(B = e_i)).
Modal (discrete A_d = V Λ V^{−1}, normalized):
MC_i = Σ_{j=1}^N ( 1 − | λ_j |^2 ) | v_{ij} |^2; for continuous-time, replace weight with −2 Re(λ_j). - S708-6 (Structural approximations):
Minimal drivers N_D = | U* | for unmatched nodes of a maximum matching;
Minimal sensors N_S = | U* | for G^T or at least the number of source SCCs. - S708-7 (Placement optimization):
Drivers: max_{U: |U| = m} f( trace(W_c(U)) ) or min trace(W_c(U)^{−1}); greedy works under submodularity conditions.
Sensors: dual of the above.
V. Metrology Workflow M7-8 (Ready → Model/Estimate → Check → Persist)
- Ready
- Fix source of A/L (analytical/identified), align Δt and RefCond; declare units.
- Choose candidate driver/sensor pools and capacity constraints (channels/cost).
- Model / Estimate
- Structural: compute maximum matching/SCC; report N_D / N_S and candidates.
- Energy: solve (discrete) Lyapunov for W_c / W_o or integrate to obtain W(T).
- Placement: greedy or MILP (small-scale) to select U / S.
- Checks
- PBH/Kalman ranks; κ(W), λ_min(W);
- Energy reachability: sample x_0, x_T and compute E_min percentiles;
- Dual forms: delta_form_ctrl / obs;
- Stability: spec(A) or ρ(A_d);
- Units/dimensions: check_dim.
- Persist / Publish
manifest.stg.ctrlobs = { A_hash, source: { struct | ident }, Δt, U, S, metrics: { rankCtrb, rankObsv, PBH_ok, κc, κo, λmin_c, AC, MC_p50, E_min_p95 }, delta: { ctrl, obs }, RefCond, method.hash }.
VI. Contracts & Assertions C70-8xx
- C70-801 (PBH pass): all eigenvalues satisfy rank([ λI − A, B ]) = N and the dual condition.
- C70-802 (Gramian PD): λ_min(W_c) ≥ τ_c, λ_min(W_o) ≥ τ_o (default τ_* = 1e−9 • scale).
- C70-803 (Conditioning): κ(W_c) ≤ κ_max, κ(W_o) ≤ κ_max (default κ_max = 1e8; beyond this, use pseudoinverse and annotate).
- C70-804 (Dual-form gap): delta_form_ctrl_p95 ≤ tol_ctrl, delta_form_obs_p95 ≤ tol_obs (suggest 1e−6 • || W ||_F).
- C70-805 (Structural lower bounds): ensure |U| ≥ N_D, |S| ≥ N_S; otherwise annotate risk and apply fallback.
- C70-806 (Stability domain): for infinite-horizon energy, require A Hurwitz or ρ(A_d) < 1.
- C70-807 (Unit consistency): check_dim( dx/dt − ( A x + B u ) ) = "[0]", check_dim( y − C x ) = "[0]".
VII. Implementation Bindings I70-8*
- I70-81 build_selection(G, policy) -> { U0, S0 }
- I70-82 structural_controllability(G) -> { N_D, drivers, scc, matching_meta }
- I70-83 gramian(A, B, C, Δt, horizon, mode) -> { W_c, W_o } (mode ∈ { lyap, integral })
- I70-84 pbh_test(A, B, C) -> { ok_ctrl, ok_obs, details }
- I70-85 energy_reach(A, B, x0, xT, T) -> E_min
- I70-86 select_actuators(A, G, m, criterion) -> U* (criterion ∈ { traceWc, minTraceInv, MC })
- I70-87 select_sensors(A, G, m, criterion) -> S* (criterion ∈ { traceWo, maxDet, observ_rank })
- I70-88 compare_forms(W_int, W_lyap) -> delta_form
- I70-89 emit_ctrlobs_manifest(results, policy) -> manifest.stg.ctrlobs
Invariants: non_decreasing(time); traceable method / hash / RefCond; delta_form_* ≤ tol_*; all check_dim pass.
VIII. Cross-References
- Graph operators & spectra: Ch. 2; kernels & diffusion energy: Ch. 4.
- Transport conservation & sources–sinks: Ch. 6 (for energy/reachability interpretation).
- Dynamics identification (source of A): Ch. 7.
- Numerical stability & time step: Ch. 9; runtime & panels: Ch. 14; manifests: Appendix C.
- Time-base & publication/integrations: companion Energy Filaments white paper (S/P/M/I).
IX. Quality & Risk Control
- SLIs/SLOs: PBH_pass_rate, rankCtrb/Obsv, κ(W_c/W_o), λ_min(W_c/W_o), AC, MC_p95, E_min_p95, delta_form_*_p95.
- Fallbacks
- PBH fails: add drivers/sensors or modify A (add damping / rescale).
- Gramian ill-conditioning: regularize W + εI, shorten horizon, normalize A.
- Structural lower bounds too high: rewire topology (edge augmentation) or blockwise control (SCC partitioning).
- Excessive energy: relocate drivers (increase AC/MC) or adjust weights/physical couplings.
- Audit: persist A/L/B/C hashes, matching & SCC certificates, PBH details, κ / λ_min, dual-form gaps, placement procedures, and seeds.
Summary
This chapter unifies structural and energy perspectives on controllability/observability: graph matching/SCCs provide lower bounds and placement priors, while PBH/Gramians provide energetic and numerical evidence. With C70-8xx contracts, dual-form gaps delta_form_*, and manifest.stg.ctrlobs.*, the path from modeling to publication is traceable and auditable.
Copyright & License (CC BY 4.0)
Copyright: Unless otherwise noted, the copyright of “Energy Filament Theory” (text, charts, illustrations, symbols, and formulas) belongs to the author “Guanglin Tu”.
License: This work is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0). You may copy, redistribute, excerpt, adapt, and share for commercial or non‑commercial purposes with proper attribution.
Suggested attribution: Author: “Guanglin Tu”; Work: “Energy Filament Theory”; Source: energyfilament.org; License: CC BY 4.0.
First published: 2025-11-11|Current version:v5.1
License link:https://creativecommons.org/licenses/by/4.0/