aiqbench analysis
This paper presents a unified Model Predictive Control framework for autonomous surface vehicles that resolves trajectory tracking and fault tolerance within a single quadratic program. By integrating post-detection thruster failures directly into the optimization, the system maintains precise path adherence even when hardware components degrade, significantly outperforming cascaded baseline architectures in live field tests. The approach demonstrates robustness through rigorous constraint handling, offering a scalable solution for over-actuated robotic systems operating under strict physical limits.
Constraint-Unified MPC for Over-Actuated Surface Vehicles with Post-Detection Fault Reconfiguration
arXiv:2609.21046v1 Announce Type: new Abstract: Choreographed aquatic performances require small autonomous surface vehicles to track precise paths under per-thruster force and rate limits, including after thruster failures. We report a deployed system in which trajectory tracking, thrust allocation, and the per-thruster force and rate limits are resolved in a single quadratic program over the per-thruster commands, with fault reconfiguration entering through one binary flag per thruster from an external detector. The system has driven a fleet in live performances on Lake Z\"urich and at the Time Space Existence 2025 exhibition in Venice. The field campaign measures 1.6 cm root mean square position error in a 10-minute hold and 4.3 cm over a 10 m square at 0.6 m/s. At 0.5 m/s, losing the front thruster increases the error to 11.6 cm, while losing the starboard-side thruster increases it to 11.5 cm. Losing two thrusters simultaneously leaves the craft tracking a 0.4 m/s square with a root mean square position error of 1.25 m. Compared to our own cascaded baseline, the unified formulation tracks the nominal square to the same few centimeters and holds station more tightly with roughly half the thruster force. The architectures separate after a thruster failure, where the unified controller stays within 24 cm of the reference path while the cascade leaves it.
Abstract:Choreographed aquatic performances require small autonomous surface vehicles to track precise paths under per-thruster force and rate limits, including after thruster failures. We report a deployed system in which trajectory tracking, thrust allocation, and the per-thruster force and rate limits are resolved in a single quadratic program over the per-thruster commands, with fault reconfiguration entering through one binary flag per thruster from an external detector. The system has driven a fleet in live performances on Lake Zürich and at the Time Space Existence 2025 exhibition in Venice. The field campaign measures 1.6 cm root mean square position error in a 10-minute hold and 4.3 cm over a 10 m square at 0.6 m/s. At 0.5 m/s, losing the front thruster increases the error to 11.6 cm, while losing the starboard-side thruster increases it to 11.5 cm. Losing two thrusters simultaneously leaves the craft tracking a 0.4 m/s square with a root mean square position error of 1.25 m. Compared to our own cascaded baseline, the unified formulation tracks the nominal square to the same few centimeters and holds station more tightly with roughly half the thruster force. The architectures separate after a thruster failure, where the unified controller stays within 24 cm of the reference path while the cascade leaves it.
15 pages, 10 figures. Accepted for the International Symposium of Robotics Research (ISRR) 2026
Robotics (cs.RO); Systems and Control (eess.SY)
arXiv:2609.21046 [cs.RO]
(or arXiv:2609.21046v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.21046 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 322 words
Related coverage

NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI
NVIDIA · May 21, 2026
Spur: Modern GPU Job Scheduling for HPC and AI Workloads
AMD · Jul 22, 2026

[AINews] OpenAI to reach AGI bar by end-2026
Latent Space · Aug 28, 2026

Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
AWS · Sep 4, 2026