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

2026-07-02

AI Summary

Existing autonomous surface vessel controllers struggle to generalize across platforms due to significant differences in hydrodynamic and actuation characteristics. This work proposes an adaptive reinforcement learning approach based on a teacher-student architecture that implicitly infers platform dynamics from interaction history, enabling zero-shot deployment of a single policy on unseen platforms without fine-tuning. Relying only on a simplified dynamical model, the method integrates partial observability modeling with randomized simulation training. Experimental validation on two real-world heterogeneous platforms demonstrates its efficacy: the proposed controller reduces position mean absolute error by up to 58% compared to non-adaptive baselines and achieves tracking accuracy comparable to platform-specific tuned controllers, marking the first demonstration of cross-platform zero-shot control without any post-deployment adaptation.

Abstract

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.

Problem

• Autonomous Surface Vehicles
• Cross-Platform Control
• Trajectory Tracking
• Generalization
• Partial Observability

Innovation

• adaptive reinforcement learning
• cross-platform control
• zero-shot deployment
• latent dynamics inference
• autonomous surface vehicles