Observer-Assisted Relative-Velocity Compensation with LPV- H infinity Robust Correction for 3D Trajectory Tracking of Underactuated Non-Minimum-Phase AUVs under Ocean Currents
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.
Rosa: We're looking at the paper: "Observer-Assisted Relative-Velocity Compensation with LPV- H infinity Robust Correction for 3D Trajectory Tracking of Underactuated Non-Minimum-Phase AUVs under Ocean Currents".
Dev: Here's the gist of the paper. This paper develops an observer-assisted control architecture specifically designed for 3D trajectory tracking of torpedotype underactuated Autonomous Underwater Vehicles (AUVs) operating in the presence of unknown ocean currents, where the vehicle exhibits non-minimum-phase sway/heave dynamics.
Rosa: The core innovation lies in combining a three-stage state–current observer with an LPV– infinity correction layer to achieve robust performance.
Dev: The control strategy is structured into three main components: 1.
Rosa: Feedforward: This component utilizes the estimated relative velocity to implement a feedforward term, which is crucial for dominant current rejection.
Paper summary: Rosa: So, to build on what we just said, let's look at the full scope of this paper. The title is "Observer-Assisted Relative-Velocity Compensation with LPV H infinity Robust Correction for three dee Trajectory Tracking of Underactuated Non-Minimum-Phase AUVs under Ocean Currents" <ref:2607.23653#pg1,Robust Correction for 3D Trajectory Tracking of Underactuated Non-Minimum-Phase AUVs>. This whole thing is about creating a new way to control those complex underwater vehicles.
Dev: It proposes this observer assisted control architecture specifically for tracking three dimensional trajectories of torpedotype underactuated AUVs that have non minimum phase sway and heave dynamics when they are in the presence of unknown ocean currents. The key claim is that by combining a three stage state current observer with an LMI certified LPV infinity correction layer, you get robust performance.
Taro: So, they're not just trying to compensate for the current; they're trying to estimate those unmeasured sway and heave velocities, and then use those estimates for a nonlinear feedforward term that handles the dominant hydrodynamic current rejection.
Rosa: And then on top of that estimation, they use this LPV infinity correction layer. This layer is designed to give certified robustness against parameter variations using an LMI formulation. The clever part is that the feedback linearizing cancellation creates a constant input matrix, which means the scheduled synthesis remains convex and avoids those problematic pairwise cross terms you usually see with parameter dependent inputs.
Dev: That’s a key technical detail because avoiding those cross terms makes the entire system design much more manageable for real time implementation. They also derived this residual level break-even law, which quantifies when that relative velocity scheduling actually improves over using just absolute velocity compensation, and they established an observer side excitation condition that determines when this whole setup is applicable.
Taro: What does this mean for the vehicle operating in the open ocean? It means you don't have to rely on perfect knowledge of the current or perfect vehicle model; you just need a good observer and a system that is robust enough to handle what's left over.
Rosa: Exactly. So, in short, it develops an observer assisted control architecture for three dee trajectory tracking of underactuated non minimum phase AUVs under unknown ocean currents by using a three stage state current observer and an LMI certified LPV infinity correction layer to achieve robust performance <ref:2607.23653#pg1,develops an observer assisted control architecture for 3D trajectory tracking of>.
Conclusion: Dev: Thinking about the title, "Observer Assisted Relative Velocity Compensation with LPV H infinity Robust Correction for three dee Trajectory Tracking of Underactuated Non Minimum Phase AUVs under Ocean Currents", it really sums up the whole approach <ref:2607.23653#pg1,Robust Correction for 3D Trajectory Tracking of Underactuated Non Minimum Phase AUVs>. It's about using an observer and a robust correction layer to deal with complex vehicle dynamics and ocean currents when you are trying to track a path.
Rosa: Yeah, and the authors are Mohammad Sabouria from the Department of Informatics, Bioengineering Robotics and Systems Engineering at University of Genoa. What they did is show that this specific combination works for these tricky AUVs. The implications are that if you're designing an AUV system, this architecture gives you a solid foundation for handling unknown ocean conditions without needing perfect knowledge of everything upfront.
Taro: For someone who only listens to the practical side, the main point is that they proved that the observer assisted relative velocity feedforward is the dominant mechanism for rejecting those currents in their simulations. That’s what matters when you're actually testing it out in a tank or a real environment.
Dev: And the LPV infinity layer provides that certified robustness, meaning you have bounds on how much things can vary and still stay within your safety margins, which is crucial for deployment where failures aren't an option. It moves the conversation from just "it works" to "it works within these defined conditions."
Rosa: So what does this mean for the broader field? It means we can start designing more complex AUVs that operate reliably in real ocean environments by using this observer assisted relative velocity feedforward as the primary tool for current rejection.
Taro: And we still have to be careful though, because the paper itself mentions that the complete closed loop stability analysis confirms practical uniform ultimate boundedness only under specific conditions like excitation and embedding margin and small gain and unsaturated cone. That means you need to keep those conditions in mind when you plan your real world testing or deployment.
Dev: Right, so the final word is this: the observer assisted relative velocity feedforward is the primary mechanism for current rejection, while the LPV infinity layer provides certified robustness, with actuator authority serving as the practical performance limit.
Mohammad Sabouria
Department of Informatics, Bioengineering, Robotics and Systems Engineering (DIBRIS), University of Genoa
eess.SY, cs.RO, cs.SY
Submitted: 2026-07-26
Updated: 2026-11-16
Comments: 31 pages, 15 figures;
Journal ref: Ocean Engineering, Vol. 370, Part 2 (2026), Article 128990
DOI: 10.1016/j.oceaneng.2026.128990
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 92/100
The gist: This paper develops an observer-assisted control architecture specifically designed for 3D trajectory tracking of torpedotype underactuated Autonomous Underwater Vehicles (AUVs) operating in the…
Key concepts
- Observer-Assisted Control
- This involves using a sophisticated estimator (the observer) that works alongside the controller. It estimates unmeasured variables like vehicle sway and the external ocean current in real-time. This estimation allows the controller to anticipate disturbances before they significantly affect the vehicle's trajectory, leading to better tracking performance.
- LPV-$\infty$ Correction Layer
- This is a control layer designed for robustness against uncertainties and parameter variations. It uses Linear Parameter-Varying (LPV) methods combined with $\infty$ gain concepts to guarantee stability even when the system's physical parameters change. It ensures the vehicle maintains performance despite modeling errors or environmental changes.
- Non-Minimum-Phase Dynamics
- This describes a challenging dynamic where the system's response to an input can be unstable or exhibit oscillations in certain directions. In this AUV, it means that controlling one movement (like surge) might inadvertently cause undesirable, unstable motions (like sway) unless carefully managed by the control strategy.
- Feedforward Compensation
- This is a proactive control technique where the system uses an estimated disturbance (the relative velocity) to immediately generate a corrective action. The paper shows this method is highly effective at rejecting ocean currents because it directly counteracts the known disturbances, significantly improving tracking accuracy.
Terminology
Summary
This paper develops an observer-assisted control architecture specifically designed for 3D trajectory tracking of torpedotype underactuated Autonomous Underwater Vehicles (AUVs) operating in the presence of unknown ocean currents, where the vehicle exhibits non-minimum-phase sway/heave dynamics. p. 0 The core innovation lies in combining a three-stage state–current observer with an LPV– infinity correction layer to achieve robust performance. p. 0
The control strategy is structured into three main components:
-
Feedforward: This component utilizes the estimated relative velocity to implement a feedforward term, which is crucial for dominant current rejection. A key finding is the derivation of a residual-level break-even law, demonstrating that after compensation based on this estimated relative velocity, the effective surge disturbance scales with the current estimation error rather than directly with the current itself. p. 1
-
Cascade Stabiliser: This provides a constant linear feedback mechanism to pre-stabilize the system dynamics, leading to a pre-stabilised Linear Parameter-Varying (LPV) matrix.
-
LPV– infinity Correction Layer: A polytopic LPV– infinity correction layer is formulated on the pre-stabilised tracking model. This layer is designed to provide certified robustness against parameter variations using an LMI formulation, and critically, the feedback-linearising cancellation yields a constant input matrix, which ensures that the scheduled synthesis remains convex and avoids problematic pairwise cross-term LMIs associated with parameter-dependent inputs. p. 1
The main contributions of this work are:
First, the development of a three-stage state–current observer to accurately estimate both unmeasured sway/heave velocities and the body-frame ocean current. Second, the derivation of a residual–level break-even law for the relative velocity feedforward mechanism. p. 1 Third, the formulation of a polytopic LPV– infinity correction layer that leverages constant input matrices from feedback linearization to maintain convex synthesis. p. 1 Fourth, a singular–perturbation stability analysis that establishes local practical uniform ultimate boundedness for the embedded observer–controller interconnection. p. 1 Finally, structured REMUS simulations confirm that the observer-assisted relative-velocity feedforward is the dominant mechanism for current rejection, while the LPV– infinity layer guarantees certified robustness and conditional scheduling margins. p. 1
The complete closed-loop stability analysis confirms practical uniform ultimate boundedness under specific conditions (excitation, embedding margin, small-gain, and unsaturated-cone), with the ultimate bound dependent on the augmented disturbance and embedding margin. p. 1 Numerical simulations demonstrate substantial performance gains, showing an 89–96% reduction in current estimation error and a significant reduction in translational residual error (from 4. p. 004 m to 0.24 m). The paper concludes that the observer–assisted relative-velocity feedforward is the primary mechanism for current rejection, while the LPV– infinity layer provides certified robustness, with actuator authority serving as the practical performance limit. p. 1
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Add an observer-assisted relative-velocity feedforward term to nonlinear feedback cancellation to provide dominant hydrodynamic current compensation. This mechanism reduces RMS tracking error from 4.038 m to 0.241 m on the constant-current helix, confirming it is the
dominant performance mechanism.
-
Implement a three-stage state-current observer that utilizes kinematic differentiation, dynamic relative-velocity estimation, and a body-frame current filter to estimate unknown ocean currents and vehicle velocities. This system enables recovery of the body frame current through
the dynamic residual route in the surge channel and through the rotational coupling.
-
Incorporate a residual-level break-even law for the relative velocity feedforward mechanism to quantify when estimated relative velocity scheduling improves over absolute velocity compensation, showing that
the effective surge disturbance depends on current-estimation error rather than with the current itself.
-
Utilize a polytopic LPV infinity control correction layer synthesized via an LMI that avoids pairwise cross terms by exploiting the
constant input matrix
from feedback linearising cancellation, providing acertified robustness mechanism.
-
Integrate a scheduling map based on estimated relative velocity and other state variables to generate the LPV-infinity gain K(phi). This allows for a
conditional robustness margin under reduced feedback authority,
showing a 9.5% RMS improvement when feedback authority is reduced to one-fifth of nominal. -
Design an online diagnostic layer that monitors the directional Gramian excitation condition (30) to identify failure mode F1, which occurs when
the transverse current estimate may not improve over v hat c = 0.
Related papers
- Industrial Load Modeling and Optimization for Market-Based Interaction with Power Systems
- Exact and Approximate Convex Reformulation of Linear Stochastic Optimal Control with Chance Constraints
- Certified Set Convergence for Piecewise Affine Systems
- Symbolic Control for Stochastic Systems via Finite Parity Games
- Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa
- Physiology-informed layered sensing for intelligent human-exoskeleton interaction