Source-linked AI summary
On Global Regulatability of Robot Manipulators by Classical PID
Cheng Zhao, Jingru Zhu, Lei Guo
TL;DR
The paper addresses whether classical PID with scalar triple gains can globally regulate robot manipulators, a question left unresolved despite PID’s practical success. It analyzes standard Euler–Lagrange systems using an energy-based Lyapunov approach and establishes a dimension-dependent result: global regulation is guaranteed in one degree of freedom but can fail for every scalar gain triple in higher dimensions.
Problem
Whether classical PID with scalar gains can globally regulate rigid robot manipulators under standard structural assumptions has remained unresolved for decades.
Method
The paper analyzes Euler–Lagrange robot manipulators under classical PID feedback and uses an energy-based Lyapunov function to study gain interactions with mechanical structure.
Results
Global regulation is guaranteed with explicit gain conditions for one-degree-of-freedom manipulators, whereas a multi-degree-of-freedom counterexample defeats every scalar PID gain triple despite globally defined closed-loop solutions.
Takeaways & Limitations
Classical scalar-gain PID has a dimension-dependent capability: it can regulate one-degree-of-freedom manipulators but has an intrinsic limitation for higher-dimensional cases.
Takeaways & Limitations
The impossibility result concerns scalar-gain PID; whether matrix-valued PID gains can globally regulate higher-dimensional manipulators remains open.
Abstract
from arXiv · showhide
A long-standing open problem in robot manipulator control is whether global regulation can be achieved by classical PID control. This paper provides an answer to this question for classical PID controllers with triple parameters (k_p,k_i,k_d) in R^3. We find and prove that for one-degree-of-freedom manipulators, the classical PID control guarantees global stability and asymptotic regulation under standard structural assumptions, and further derive explicit quantitative design conditions for the PID gains. However, for multi-degree-of-freedom cases, we can construct a robot manipulator satisfying the same structural assumptions for which no choice of PID gains (k_p,k_i,k_d) can achieve global asymptotic regulation. These results provide a fundamental understanding of the abovementioned open problem, revealing both the fundamental capability and intrinsic limitation of the classical PID control for robot manipulator dynamics.
1 Introduction
PID control remains widely used because of its simple structure, low implementation cost, and engineering effectiveness, yet its global regulatability for robot manipulators has been unresolved. This paper gives a dimension-dependent characterization: classical scalar-gain PID succeeds in one degree of freedom but has intrinsic limitations in higher dimensions.
- Motivation: PID control remains widely adopted because it combines simple structure, low implementation cost, and engineering effectiveness.It continues to serve as a fundamental design paradigm in industrial control systems.
- Related work: Systematic PID theory has established global stabilization conditions and admissible gain regions for several classes of second-order nonlinear systems.These results support PID design beyond empirical trial and error.
- Related work: Broad nonlinear-system PID analyses use limited prior information such as Lipschitz bounds and growth-rate conditions, but may underexploit mechanical structures.Relevant structures include conservative vector fields, energy properties, and Lagrangian formulations.
- Related work: Prior robot-control results show that PD control can globally stabilize frictionless manipulators, while arbitrary desired configurations generally require gravity compensation or additional structure.Without gravity compensation, the desired configuration must itself be an equilibrium for pure PD feedback.
- Open problem: The unresolved question is whether standard linear PID with scalar gains can globally regulate robot manipulators without changing the controller structure.Nonlinear integral actions, saturation, and other modifications recover regulation in related work but do not answer this classical-PID question.
- Contribution: For one-degree-of-freedom manipulators, the paper proves global stability and asymptotic regulation under standard structural assumptions and derives explicit PID gain conditions.The proof uses an energy-based Lyapunov function combining augmented error states with gravitational potential energy.
- Contribution: For n ≥2 degrees of freedom, counterexamples satisfy standard rigid-manipulator assumptions, yet every scalar PID gain triple fails to achieve regulation.The failure occurs despite globally defined closed-loop solutions, establishing a limitation of the scalar-gain architecture rather than loss of existence.
2 Problem Formulation
The paper formulates global regulation for Euler–Lagrange robot manipulators controlled by classical PID feedback with one scalar gain triple shared across all degrees of freedom. The objective requires globally defined solutions and convergence of position and velocity errors to zero from arbitrary initial conditions, while the answer depends on dimension.
- System model: The manipulator is modeled by standard Euler–Lagrange dynamics with generalized coordinates, velocities, accelerations, inertia, Coriolis and centrifugal terms, gravity, and control input.The dynamics are written as M(q)¨q + C(q, ˙q) ˙q + g(q) = τ.
- Structural assumptions: The structural assumptions include standard rigid-manipulator properties and an additional symmetry condition on the Coriolis mapping.The assumptions are parameterized by positive constants m0, m1, Lc, and Lg.
- Regulation objective: The control objective is to drive the manipulator to an arbitrary constant desired configuration from arbitrary initial conditions.The position error is used to specify the regulation objective.
- Controller architecture: The analysis focuses on classical PID feedback with scalar gains (kp, ki, kd) ∈R3 and an integral state ξ(t) ∈Rn.The same gain triple is applied to every degree of freedom; matrix-valued gains are outside the stated architecture.
- Regulation definition: Global regulation requires closed-loop solutions for all t ≥0 and convergence e(t) →0 and ˙e(t) →0 as t →∞.An admissible PID gain is one gain vector satisfying these requirements for every initial condition and desired configuration.
- Problem statement: Global regulation by classical PID for rigid manipulators satisfying standard assumptions remained unresolved despite the controller’s practical success.The paper identifies the answer as intrinsically dimension-dependent: sufficient in one degree of freedom but not generally sufficient for two or more.
3 The Main Results
For one-degree-of-freedom manipulators, explicit scalar PID gain conditions ensure global regulation, while for multi-degree-of-freedom systems scalar PID gains cannot guarantee it under the same structural assumptions.
- One-degree-of-freedom case: The gain condition depends only on uniform inertia bounds m0, m1 and the global gravity slope bound Lg, not exact expressions of M, C, or g.This provides a model-light gain-selection rule based on structural bounds.
- One-degree-of-freedom case: For n = 1, Theorem 1 establishes global regulation with exponential rate for every desired setpoint and initial condition when gains belong to the admissible set K1.The result assumes Assumption 1 and classical PID gains.
- One-degree-of-freedom case: Positive ki and kd may be selected arbitrarily, after which kp can be chosen sufficiently large to satisfy the gain condition.In particular, kd may be arbitrarily small because derivative feedback supplies natural dissipation through the Euler–Lagrange energy structure.
- One-degree-of-freedom case: The sufficient threshold on kp represents a gravity-dominance requirement, with inertia bounds influencing the threshold through the Lyapunov analysis.The proportional feedback must dominate the worst-case variation of the gravity force.
- Multi-degree-of-freedom case: For n ≥ 2, the same structural assumptions are insufficient: there exists a robot manipulator that fails global regulation for every scalar gain triple in R^3.The counterexample uses complete trajectories that do not converge to the desired setpoint, so the impossibility is not caused by finite escape.
- Multi-degree-of-freedom case: The higher-dimensional impossibility result is specific to scalar-gain classical PID; whether matrix-valued PID gains can achieve global regulation remains open.This identifies a scope boundary rather than resolving all PID architectures.
4 Proof of the Main Results
The proofs establish global exponential stability and asymptotic regulation for one-degree-of-freedom manipulators under explicit PID gain conditions, then construct a higher-dimensional counterexample defeating every scalar-gain PID controller.
- Theorem 1 proof: The one-degree-of-freedom proof reformulates the closed loop using augmented coordinates and constructs an energy-based Lyapunov function.The analysis uses scalar inertia and Coriolis terms, bounded gravity variation, and a quadratic-form matrix P.
- Theorem 1 proof: The Lyapunov function is positive definite and radially unbounded when the gain conditions make the associated matrix positive definite.The proof bounds the gravity potential contribution and applies principal-minor and Schur-complement arguments.
- Theorem 1 proof: A uniformly positive-definite dissipation matrix yields global exponential stability, which implies convergence of the regulation error and its derivative.The derivative satisfies V̇ ≤−β0c∥z∥2 < 0 away from the equilibrium.
- Theorem 2 proof: For n≥2, the proof constructs a smooth Euler–Lagrange manipulator with bounded gravity and uniformly bounded, positive-definite inertia for which every scalar PID gain triple fails.The argument separates all gain triples into mutually exclusive cases and shows global existence does not prevent regulation failure.
- Theorem 2 proof: The counterexample rules out gain cases through persistent error, nonvanishing acceleration, and an unstable linearization, while its trajectories remain forward complete.Thus the impossibility is not caused by finite escape or loss of solution existence.
- Theorem 2 proof: The proof’s implication from vanishing error and derivative to vanishing integral state depends on mechanical structure and fails for general third-order systems.The argument uses moving-average convergence together with η̇=e to obtain pointwise convergence.
5 Conclusion
The paper resolves global regulatability for classical scalar PID in a degree-of-freedom-dependent way. It establishes explicit one-degree-of-freedom guarantees while constructing a multi-degree-of-freedom counterexample where every scalar gain triple fails.
- The paper addresses the long-standing open problem of global regulatability for classical PID control of robot manipulators.
- In one degree of freedom, explicit PID gain inequalities guarantee global exponential regulation for every initial condition and setpoint.
- For dimension n ≥2, the same structural assumptions do not imply global regulatability by scalar PID gains.
- A smooth mechanical counterexample shows that every scalar gain triple fails for some initial condition and setpoint.
- Multi-degree-of-freedom regulation therefore requires additional structure, richer gain matrices, model-dependent compensation, or modified nonlinear integral actions.