Skip to content

Compare methods

Compare source explanations, strengths, limitations and references for up to four methods.

7 results
Criterion Active Disturbance Rejection Control Model Predictive Control Adaptive Control
Overview
  • Active Disturbance Rejection Control (ADRC) is a control method that estimates and cancels disturbances and model uncertainties in real-time using an extended state observer (ESO), treating them as additional 'disturbance states'. Developed by Han and popularized by Gao, ADRC achieves remarkable robustness without requiring precise plant models, making it practical for real-world systems with significant uncertainty and disturbances.
  • Model Predictive Control (MPC) is an advanced control strategy that uses an explicit process model to predict future system behavior over a finite horizon and solves an optimization problem at each control step. First formalized by Richalet et al. in 1978, MPC has become the dominant approach in process control industries, from chemical plants to autonomous vehicles, because it naturally handles constraints and can optimize multiple objectives simultaneously.
  • Adaptive Control is a control strategy that adjusts controller parameters in real-time based on online system identification to maintain performance despite changing plant dynamics or uncertain parameters. Pioneered by Astrom and Wittenmark, adaptive control enables robust operation in time-varying environments, from aircraft with fuel depletion to industrial systems with aging components.
Use and interpretation
  • Use ADRC when the plant model is inaccurate, disturbances are substantial, and you want simple control law without detailed model. Ideal for industrial applications (motors, pumps, heating), robotics with uncertain payloads, and systems with time-varying parameters. Avoid ADRC if the disturbance is too fast for the observer to track or if very low steady-state ripple is required.
  • Use MPC when you need to handle constraints explicitly (e.g., valve limits, safety bounds), manage multiple correlated variables (multivariable control), or when disturbances and non-linearities are significant. It is most effective for systems with slow to moderate dynamics where computation time allows for real-time optimization. Typical applications include chemical process control, HVAC systems, power plants, and autonomous driving. Avoid MPC if your system demands microsecond control rates or if a high-fidelity process model is unavailable.
  • Use adaptive control when system parameters vary significantly over time (aging, temperature, load changes), when parameters are unknown but system structure is known, and when you need guaranteed stability despite parametric uncertainty. Ideal for aircraft control (fuel depletion, CG shift), industrial process control (catalyst aging), and renewable energy systems (wind/solar varying). Avoid adaptive control if parameters change faster than identification can track, if stability margins must be very conservative, or if a fixed robust controller suffices.
Strengths
  • Robust to model uncertainty; treats uncertainty as estimated disturbance rather than designing for worst-case.
  • Simple control law (often PD-like); minimal tuning compared to PID or H-infinity.
  • Fast disturbance rejection; real-time estimation enables rapid response.
  • Extends to MIMO and nonlinear systems naturally.
  • Practical implementation; no need for detailed model identification.
  • Naturally handles input and state constraints through explicit optimization.
  • Provides multivariable control, coordinating multiple inputs and outputs automatically.
  • Robust to disturbances and model mismatch because of the receding horizon feedback loop.
  • Allows incorporation of future reference trajectory or set-point changes.
  • Easy to tune (cost function weights) compared to traditional PID tuning.
  • Maintains performance as system parameters drift; no manual retuning required.
  • Faster response to parameter changes than fixed robust controllers.
  • Reduces steady-state error without over-designing for worst-case.
  • Enables discovery of unknown parameters during operation.
  • Naturally incorporates measurement data for real-time learning.
Limitations
  • Requires accurate output measurement; observer sensitivity to noise can limit performance.
  • Observer bandwidth limited by measurement noise; cannot estimate arbitrarily fast disturbances.
  • High observer gain can amplify measurement noise; trade-off between disturbance estimation and noise rejection.
  • Tuning ESO parameters (observer gains) requires experience; less systematic than classical control methods.
  • Theoretical analysis more complex; stability and convergence guarantees less mature than classical methods.
  • Computationally expensive; solving the optimization problem at each step may be infeasible for very fast systems.
  • Requires an accurate process model; poor model accuracy degrades performance significantly.
  • Nonlinear MPC is non-convex and may get stuck in local minima.
  • Stability and feasibility guarantees are non-trivial, especially for constrained problems.
  • Identification phase requires persistent excitation (sufficiently varied input); slow in quiet environments.
  • Parameter estimation convergence can be slow; transient instability possible during adaptation.
  • Requires accurate measurement of outputs; noise and delays degrade identification.
  • Stability proofs are complex; adaptive systems can be unstable despite individual components being stable.
  • Adaptation rate must be slower than control loop; conflicts with fast tracking requirements.
Common pitfalls
  • Setting observer gains too high; leads to noise amplification and instability.
  • Using ADRC for systems with fast unmodeled dynamics; ESO bandwidth may be insufficient.
  • Ignoring measurement quality; poor sensor signal degrades observer and overall performance.
  • Assuming all disturbances are observable; if disturbance does not directly affect measured output, ESO cannot estimate it.
  • Setting the prediction horizon too short leads to myopic, suboptimal control; too long increases computation.
  • Failing to tune the cost function (e.g., weighting between error and control effort) leads to sluggish or aggressive responses.
  • Using an oversimplified or inaccurate model that does not capture system dynamics results in poor real-world performance.
  • Over-aggressive adaptation gain; causes instability when estimates overshoot true values.
  • Insufficient persistent excitation; parameters fail to converge and adaptive control does not improve performance.
  • Using adaptive control in non-stationary environment; identification never converges if plant keeps changing.
  • Ignoring estimation error; uncertainty in parameter estimates degrades closed-loop performance.
Introduced by
  • Jingquan Han
  • Jacques Richalet
  • Karl J. Astrom
Year
  • 2009
  • 1978
  • 1983