Adaptive Tracking Control of Euler-Lagrange Systems with Time-Varying State and Input Constraints

This paper proposes an adaptive control framework for Euler-Lagrange systems that guarantees user-defined time-varying state and input constraints under uncertainties and disturbances by integrating a time-varying barrier Lyapunov function with a saturated control law, supported by an offline feasibility condition and validated through real-time helicopter experiments.

The Gist

Imagine you are driving a high-performance race car, but with a twist: the road ahead is constantly changing. Sometimes the speed limit drops to 10 mph because of a school zone; other times, the road narrows, and you can't turn the steering wheel more than a few degrees. To make things harder, the car's engine is a bit unpredictable (it might sputter or surge), and there are random gusts of wind pushing you off course.

Your goal is to drive exactly where a "ghost driver" (the reference trajectory) tells you to go, but you must never hit the walls, never exceed the speed limit, and never twist the steering wheel so hard that it breaks.

This paper presents a new "smart autopilot" system designed to handle exactly this kind of chaotic, changing situation. Here is how it works, broken down into simple concepts:

1. The Problem: The "Moving Goalposts"

Most old-school controllers are like a driver who only knows the rules for a straight, empty highway. They assume the speed limit is always 100 mph and the road is always wide.

• The Reality: In real life (like with drones, surgical robots, or helicopters), the "safe zone" changes. A helicopter might need to fly very slowly and carefully when hovering near a person, but can fly faster when far away.
• The Risk: If the controller tries to correct a mistake too aggressively, it might push the engine into "redline" (saturation), causing the system to crash or become unstable.

2. The Solution: The "Invisible, Stretching Bubble"

The authors created a control system that wraps the machine in an invisible, stretchy bubble.

• The Bubble (Constraints): This bubble represents the safe limits. It can shrink, expand, or move around.
• The Stretchy Walls (Time-Varying): Unlike a rigid cage, this bubble changes shape over time. If the task requires a tight maneuver, the bubble shrinks. If the task is relaxed, it expands.
• The Magic: The controller is designed so that the machine physically cannot touch the walls of this bubble. It acts like a force field that gently pushes the machine back toward the center if it gets too close to the edge.

3. The Secret Sauce: "Offline" Safety Checks

Usually, to keep a robot safe in a changing environment, you need a supercomputer to solve complex math problems while the robot is moving (Real-time Optimization). This is slow and risky if the computer gets overwhelmed.

This paper introduces a clever trick: The Pre-Flight Checklist.

• Before the robot even starts moving, the system runs a check (an "offline feasibility condition").
• It asks: "Is the bubble we want to create actually possible to stay inside, given the engine's strength and the wind?"
• If the answer is Yes, the system locks in the plan. If the answer is No, it tells the user: "Hey, you asked for a bubble that is too small for this engine. Make the bubble bigger or the engine stronger."
• The Benefit: Once the robot starts, it doesn't need to do complex math. It just follows the pre-approved plan, making it fast and reliable.

4. How It Handles Mistakes (Adaptive Learning)

The system knows the machine isn't perfect. Maybe the helicopter is heavier than expected, or the wind is stronger.

• The "Self-Correcting" Mechanism: The controller has a built-in "learning" feature. If the wind pushes the helicopter, the controller notices the error and quietly adjusts its internal settings to compensate, without ever breaking the safety bubble.
• The "Brake" (Saturation): If the controller tries to push too hard and hits the engine's maximum limit, the system has a special "brake" that prevents the engine from burning out. It gracefully handles the limit instead of panicking.

5. The Real-World Test: The Helicopter

To prove this works, the authors tested it on a 2-DoF Helicopter (a small model that can pitch up/down and yaw left/right).

• The Challenge: They made the helicopter follow a wiggly path while constantly changing the speed limits and turning limits.
• The Result: The helicopter stayed perfectly within the "stretchy bubble." It never hit the walls, never broke the engine limits, and tracked the path smoothly, even with wind and model errors.

Summary Analogy

Think of this system as a dance instructor for a robot.

• Old methods: The instructor shouts, "Move fast!" or "Stop!" based on a fixed rulebook. If the robot hits a wall, it crashes.
• This paper's method: The instructor holds a stretchy, glowing hoop around the robot. The hoop changes size and shape based on the music (the task). The instructor has already checked the map to ensure the robot can dance inside that hoop without tripping. If the robot stumbles, the instructor gently nudges it back in, ensuring it never touches the floor (the constraints) or breaks its own legs (the actuators).

In short: This paper gives engineers a way to tell a robot, "Do exactly what I want, but stay strictly within these moving safety lines, and I promise you won't break anything."

Technical Summary

Here is a detailed technical summary of the paper "Adaptive Tracking Control of Euler-Lagrange Systems with Time-Varying State and Input Constraints."

1. Problem Statement

The paper addresses the control of uncertain Euler-Lagrange (E-L) systems (e.g., robotic manipulators, aerial vehicles) subject to:

• Parametric Uncertainties: Unknown constant parameters in the dynamic model.
• Bounded Disturbances: External forces or modeling errors.
• Time-Varying Constraints: Both state constraints (position and velocity must stay within time-varying envelopes) and input constraints (control torque/voltage must stay within time-varying limits).

The Core Challenge:
Existing methods often handle static constraints or time-varying constraints separately.

• Optimization-based methods (MPC) are computationally expensive and require accurate models.
• Barrier Lyapunov Functions (BLFs) handle static constraints well but often ignore actuator saturation, leading to instability when control effort exceeds physical limits.
• Prescribed Performance Control (PPC) and Funnel Control (FC) handle time-varying error bounds but typically fail to account for actuator saturation or require modifying the reference trajectory online when saturation occurs.

The goal is to design an optimization-free adaptive controller that guarantees tracking performance while strictly enforcing both time-varying state and input constraints without altering the reference trajectory.

2. Methodology

The proposed framework integrates a Time-Varying Barrier Lyapunov Function (TVBLF) with a saturated control law and adaptive parameter estimation.

A. System Dynamics and Assumptions

The system is modeled as:
$$M(q)\ddot{q} + V_m(q, \dot{q})\dot{q} + G_r(q) + F_d(\dot{q}) = \tau$$
Standard E-L properties are assumed (symmetric positive-definite inertia, linear parameterizability, skew-symmetry of $\dot{M}-2V_m$).

B. Constraint Transformation

To enforce constraints on the plant states ($q, \dot{q}$), the authors transform them into constraints on the tracking errors ($e = q - q_d$) and a filtered tracking error ($r = \dot{e} + \alpha e$).

• Assumption 1: The reference trajectory and its derivatives are bounded within the safe region.
• Lemma 1: Establishes a mathematical relationship proving that if the filtered error $r(t)$ remains within a specific time-varying bound $\phi_r(t)$, then the position and velocity errors ($e, \dot{e}$) will automatically satisfy their respective time-varying bounds.

C. Control Design

The control input $\tau(t)$ is designed in two steps:

1. Auxiliary Control ($\tau_a$): An adaptive control law is derived using the TVBLF to stabilize the error dynamics. It includes a term to counteract parametric uncertainties ($Y\hat{\theta}$) and a damping term ($Kr$).
2. Saturation Logic: The actual control input is the saturated version of $\tau_a$:
   $$\tau(t) = \begin{cases} \tau_a(t) & \text{if } \|\tau_a(t)\| \leq \phi_\tau(t) \\ \phi_\tau(t) \frac{\tau_a(t)}{\|\tau_a(t)\|} & \text{if } \|\tau_a(t)\| > \phi_\tau(t) \end{cases}$$
   This ensures the input never exceeds the time-varying limit $\phi_\tau(t)$.

D. Stability Analysis

A composite Lyapunov function $V$ is constructed, combining the TVBLF (for state constraints) and a parameter estimation error term.

• Adaptive Law: A projection-based update law is used for parameter estimation to ensure boundedness.
• Feasibility Condition (C1): A critical contribution is the derivation of an offline-verifiable feasibility condition. This inequality ensures that the available control authority ($\phi_\tau$) is sufficient to handle the worst-case dynamics, uncertainties, and the rate of change of the constraints ($\dot{\phi}_r$).
  • If Condition C1 is met, the paper proves that all closed-loop signals are bounded, and the system never violates the constraints (forward invariance).
  • If the saturation error vanishes after some time, asymptotic tracking ($e \to 0$) is recovered.

3. Key Contributions

1. Optimization-Free Framework: The controller enforces time-varying state and input constraints without solving real-time optimization problems (unlike MPC), making it suitable for systems with limited computational power.
2. Offline Verifiable Feasibility Condition: The authors derive a sufficient condition (C1) that allows engineers to verify a priori whether a specific set of time-varying constraints is achievable given the system's physical limits and uncertainties.
3. Unified Handling of Constraints: Unlike previous works that treat state and input constraints separately or modify the reference trajectory upon saturation, this method guarantees constraint satisfaction for both simultaneously while maintaining the original tracking objective.
4. Generalization of PPC/FC: The framework subsumes Prescribed Performance Control (PPC) and Funnel Control (FC) as special cases by choosing specific forms for the time-varying bounds (Performance Functions).

4. Experimental Results

The method was validated on a Quanser 2-DoF Helicopter platform.

• Setup: The helicopter's pitch and yaw angles were controlled to track a sinusoidal reference trajectory.
• Constraints: Time-varying bounds were defined for position error, velocity error, and control torque using performance functions (decaying envelopes).
• Performance:
  • The pitch and yaw angles successfully tracked the reference.
  • State Constraints: The position and velocity errors remained strictly within the predefined time-varying envelopes.
  • Input Constraints: The control torques remained within the time-varying input limits, even during transient phases where errors were large.
  • Robustness: The system handled parametric uncertainties and external disturbances without violating safety limits.

5. Significance and Impact

• Safety-Critical Applications: This work is highly relevant for safety-critical systems (surgical robots, drones, autonomous vehicles) where violating state or input limits can cause physical damage or loss of control.
• Practical Viability: By avoiding real-time optimization, the method is computationally efficient and suitable for embedded hardware.
• Design Flexibility: The "offline verifiable" nature of the feasibility condition provides a rigorous design tool. Engineers can define performance requirements (how fast the system must converge) and immediately check if the hardware can physically achieve them without trial-and-error.
• Theoretical Advancement: It bridges the gap between adaptive control, barrier Lyapunov functions, and input saturation, offering a rigorous solution to a problem that has historically required trade-offs between performance and safety.