Critique: Coulomb Friction Violation of Drift Invariance

Critique and response context for Coulomb Friction Violation of Drift Invariance in AffineDrift’s control-affine golf-swing framework.

Critique: Coulomb Friction Violation of Drift Invariance

Summary of Concern

The central proof of Drift Invariance (Proposition 1) asserts that the drift vector field \(f(x)\) is independent of the input \(u\) (\(\nabla_u f(x) \equiv 0\)). This proof relies on the assumption that all passive forces (included in \(h(x)\)) depend only on state \((q, \dot{q})\).

However, real mechanical systems contain Coulomb friction at joints, where \(\tau_{fric} = \mu F_N \operatorname{sgn}(\dot{q})\). The normal force \(F_N\) is a component of the constraint force vector, which depends explicitly on the applied acceleration and thus the input torque \(u\) (via the Constraint Jacobian). Therefore, in the presence of realistic dry friction, the passive resistance \(f(x)\) becomes a function of \(u\), violating the affine structure \(\dot{x} = f(x) + g(x)u\).

Location

  • Page: articles/theory-part3.qmd
  • Section: Drift Invariance and Input Constraints (Proposition 1)
  • Claim: “The drift vector field \(f(x)\) is invariant with respect to the control input \(u\)… friction terms are linear in velocity (viscous).”

Nature of the Issue

  • Modeling Assumption Failure: The proof assumes only viscous damping (\(D\dot{q}\)). It ignores dry friction, which is significant in loaded biological joints and mechanical linkages.
  • Mathematical Inconsistency: If Coulomb friction is present, the system is no longer control-affine. The decomposition \(\tau_{input} = \tau_{total} - \tau_{drift}\) fails because \(\tau_{drift}\) cannot be calculated without knowing \(\tau_{input}\).

Why This Is a Problem

  1. Exactness Claim: The paper claims the decomposition is “analytically exact” for the model. This is only true for a specific, friction-simplified model.
  2. Magnitude: In high-load scenarios (like the golf downswing where joint reaction forces are huge), the friction variation due to input-induced normal force loading could be non-negligible. The “passive” resistance increases as the player pushes harder.
  3. Causal Leakage: A portion of the “Input” (effort) is instantly consumed by the “Drift” (friction increase) it creates. The ZTCF (where \(u=0\)) would underestimate the friction present in the actual swing, making the “Drift” look more efficient than it is.

Evidence / References

  • Featherstone, R. (2008). Rigid Body Dynamics Algorithms. (Constraint forces depend on applied forces).
  • Pain, M. T. G., & Challis, J. H. (2006). “The influence of soft tissue movement on ground reaction forces…” (Joint reaction forces in biomechanics).

Severity

  • Medium: It limits the “Exactness” claim but likely doesn’t destroy the macro-level utility of the framework for swinging motions (where inertial forces dominate friction).

Suggested Remedies

  1. Explicit Exclusion: Add “Coulomb Friction” to the list of “Limitations” (alongside Aerodynamics).
  2. Justification: Argue that for high-speed ballistic motions, inertial terms (\(M\ddot{q} \sim \omega^2\)) dominate frictional terms, making the viscous approximation acceptable.
  3. Refined Claim: Change “Structurally Independent” to “Structurally Independent (under the assumption of viscous-only damping)”.