Dynamical System-Based Imitation Learning and Neuroadaptive Control for Trajectory Recovery in Autonomous Ships

Yeyson A. Becerra-Mora; José Ángel Acosta

eess.SYAug 25, 2026

Dynamical System-Based Imitation Learning and Neuroadaptive Control for Trajectory Recovery in Autonomous Ships

Authors: Yeyson A. Becerra-Mora, José Ángel Acosta

Organizations: Dept. Ingenieria de Sistemas y Automatica, University of Seville, Sevilla, 41092, Spain · Dept. of Electronic Engineering, CUN, Bogota, 111711, Colombia

Abstract

Repetitive maritime operations can be effectively learned using the Imitation Learning (IL) paradigm, which transfers human expertise directly to Unmanned Surface Vehicle (USV) control systems. Dynamical Systems (DS) are widely used to model non-linear human demonstrations while offering inherent stability guarantees. However, real-world execution under persistent marine perturbations reveals a critical trade-off: standard DS-based IL approaches prioritize global target convergence at the expense of localized trajectory reproduction fidelity. To address this limitation, we present a hybrid learning-control architecture that integrates a DS-based IL reference generator with a neuroadaptive controller. Our approach introduces a control action that drives the USV back to the demonstrated path following exogenous disturbances, enabling dynamic human-like reactive alignment-termed behavioral tracking. The proposed methodology is validated using the Marine Systems Simulator (MSS) toolbox. Simulation results confirm that the framework generalizes complex maneuvering tasks while substantially improving trajectory tracking fidelity under disturbances compared to alternative control strategies.

Figures & tables

Figure 1: IL architectures comparison: top-left, learning references for human-like responsiveness; bottom-left, demonstration references for classical trajectory tracking control.

Figure 3: Demonstrated trajectories (black dots) and estimated trajectories (blue lines).

Panel of Figure 4 — Figure 4: Comparison of references. Demonstration-based reference (a) and learning-based reference (b).

Type of trajectory	Controllers
	NA	MRAC	PD
A	0.0111	0.0289	0.0255
B	0.0368	0.0561	0.0575
C	0.0490	0.0913	0.0837

Table 2: Trajectory tracking error with ideal conditions (in Nautical miles)

Figure 7: Heading tracking with noisy observations.

Figure 8: Heading tracking with noisy observations.

Type of trajectory	Controllers
	NA	MRAC	PD
A	0.0231	0.1561	0.1839
B	0.0530	0.4268	0.4386
C	0.0869	0.2552	0.2276

Table 3: Trajectory tracking error with noisy observations (in Nautical miles)

Disturbance Rejection

See all 13 papers

Cross-Platform Control for Autonomous Surface Vehicles via Adaptive Reinforcement Learning
Jul 2, 2026Ruiheng Jiang, Thomas Bi, Raffaello D'Andrea +1Robotic Control Robot Trajectory Tracking
Institute for Dynamic Systems and Control, ETH Zurich, Switzerland.
Meta-Learning for Rapid Adaptation in Reference Tracking of Uncertain Nonlinear Systems
May 21, 2026Jiaqi Yan, Ankush Chakrabarty, Niklas Schmid +2Robot Trajectory Tracking Meta-Learning
School of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, P. R. China · Mitsubishi Electric Research Laboratories, Cambridge, MA 02139, USA · Automatic Control Laboratory, ETH Zurich, Switzerland +1
Sim-to-Real RL for ASVs using SysID
Oct 8, 2026Cody Sheltraw, Tsimafei Lazouski, Maani Ghaffari +1Sim-to-Real Transfer Robotics Simulation
University of Michigan, Ann Arbor, MI 48109, USA.

Jul 2, 2026J/K move · Enter open · S save

Cross-Platform Control for Autonomous Surface Vehicles via Adaptive Reinforcement Learning

Ruiheng Jiang, Thomas Bi, Raffaello D'Andrea, Aswin Ramachandran

Institute for Dynamic Systems and Control, ETH Zurich, Switzerland.

Robotic Control Robot Trajectory Tracking Reinforcement Learning Sim-to-Real Transfer Cross-Embodiment Robot Learning

Open paper AskPDF arXiv

Autonomous surface vehicles vary widely in hydrodynamic and actuation characteristics, yet most controllers are designed for single-platform deployment. We present an adaptive reinforcement learning approach for trajectory tracking that enables zero-shot cross-platform deployment using a single policy. Since the deployment platform's dynamics are unknown to the policy, we address cross-platform generalization with the standard partial-observability approach of conditioning on interaction history, employing a teacher-student architecture in which a learned module infers a latent representation of the platform dynamics. The policy is trained in simulation under randomized vessel dynamics and is deployed zero-shot to two real-world platforms without any fine-tuning, despite relying on a simple analytical dynamics model rather than a high-fidelity hydrodynamic simulator. In real-world experiments on two different platforms, the adaptive policy outperforms non-adaptive learning-based baselines by up to 58% in position mean absolute error while approaching the tracking accuracy of a platform-specific tuned controller.

Abstract

Disturbance Rejection

Explore similar work

Cross-Platform Control for Autonomous Surface Vehicles via Adaptive Reinforcement Learning

Meta-Learning for Rapid Adaptation in Reference Tracking of Uncertain Nonlinear Systems

Sim-to-Real RL for ASVs using SysID