You know how a good DJ runs a mixing board — each channel's fader controls one instrument, and you adjust bass, treble, and volume per channel without worrying that turning up the vocals will blow out the drums? That's exactly what robotics engineers have been doing with PID controllers on multi-joint robots for decades: tuning each joint's proportional, integral, and derivative gains independently, even though the joints are physically coupled through inertia, gravity, and Coriolis forces. Everyone does it. Nobody had the proof that it works under the conditions that actually matter. This paper provides that proof. The committed claim: for a class of uncertain MIMO nonlinear systems with channel-varying relative degrees, an extended PID (EPID) controller tuned independently per channel guarantees semiglobal stability and setpoint regulation, provided the unknown nonlinearities satisfy specified growth conditions. The 'extended' part means some channels use higher-order derivative terms — if a joint has relative degree 3 instead of 2, you add a third-derivative term. The key insight is that the admissible parameter set can be constructed from known bounds on initial states, uncertainties, and the reference signal, without requiring exact knowledge of the plant dynamics. Where does this sit on the ladder? The classical baseline is computed-torque control (also called feedback linearization), which requires a full dynamic model of the robot. Model-free PID has always been the pragmatic alternative, but theoretical guarantees were limited to single-channel or fully actuated systems with uniform relative degree. Recent work by Zhao and Guo (2024) extended PID theory to SISO nonlinear systems; this paper generalizes to MIMO with non-uniform relative degrees across channels. It doesn't beat computed-torque on performance — it beats it on assumptions, requiring no model at all. The architecture is classical control theory, not learning-based. The proof technique constructs a Lyapunov-like argument channel by channel, then assembles the overall stability guarantee. The method leans on the structural assumption that the system has a well-defined vector relative degree — a property that most serial-link robot manipulators satisfy but that parallel mechanisms or flexible-link robots may not. The growth conditions on the nonlinearities are the load-bearing assumption: they must not grow faster than certain polynomial bounds relative to the state. Integrity is strong for a theory paper. The results are mathematical proofs, not simulations cherry-picked to look good. The paper explicitly states the conditions under which the guarantees hold and does not overclaim. However, there are no experimental demonstrations — no hardware, no simulation, no numerical examples in the abstract. This is pure theory. The validation is the proof itself, which means the grading is internal: the math either holds or it doesn't, but we don't see how the required growth conditions fare against real robot dynamics. The milestone question is concrete: can you specify the admissible PID parameter set for a real 6-DOF industrial manipulator (say a KUKA or Fanuc arm) with quantified uncertainty bounds, and demonstrate that the resulting controller performs acceptably in hardware? That's the bridge from proof to practice. The gap is probably 2-4 years of follow-on experimental work. The obvious experiment not run is a numerical simulation on a standard benchmark robot (2R planar arm, SCARA, or 6-DOF manipulator). Every controls paper in this space includes one. The honest read: this is a theory-first paper from authors whose prior work (Zhao & Guo 2024) is also theory-first. They're building the mathematical edifice first, simulations second. Likely saving the demonstration for a follow-on paper or a journal version — option (c).
