Underwater Robotics Gets a Budget Boost

by

Author: Denis Avetisyan

Researchers have unveiled a new, affordable multi-robot platform to advance studies in autonomous navigation and underwater mapping.

A dynamically reconfigurable collective of underwater robots demonstrates adaptability through coordinated movement, enabling complex behaviors and distributed task execution in challenging aquatic environments.
A dynamically reconfigurable collective of underwater robots demonstrates adaptability through coordinated movement, enabling complex behaviors and distributed task execution in challenging aquatic environments.

This paper details the CoUGARs system—a low-cost, ROS 2-based platform for acoustic cooperative localization and multi-agent systems research utilizing autonomous underwater vehicles.

Real-world validation of multi-agent autonomy remains hampered by the substantial financial and logistical barriers to field testing. This paper details the development of the Configurable Underwater Group of Autonomous Robots (CoUGARs), a low-cost, customizable platform designed to accelerate research in acoustic cooperative localization and underwater navigation. The system, built with commercially-available and 3D-printed components for a base cost under $3,000 USD, integrates a doppler velocity log, ultra-short-baseline acoustic array, and containerized software stack for robust state estimation and communication. Will this accessible platform democratize multi-agent research and facilitate breakthroughs in underwater autonomy and mapping?

Navigating the Depths: The Challenge of Underwater Localization

Traditional localization methods for autonomous vehicles rely heavily on the Global Positioning System (GPS), a dependency that presents a critical limitation for underwater robotics, as GPS signals are rapidly attenuated beneath the water’s surface. Accurate and reliable underwater positioning is paramount for applications ranging from environmental monitoring to critical infrastructure inspection and coordinated multi-robot operations. The absence of robust localization capabilities restricts the complexity and scope of underwater missions. Progress in this field necessitates the development of self-contained positioning systems capable of operating effectively in this challenging environment.

The pose graph estimates the vehicle's trajectory by incorporating measurements from diverse sources, including a prior, visual odometry (DVL), inertial measurement unit (IMU), depth sensors, and GPS.
The pose graph estimates the vehicle’s trajectory by incorporating measurements from diverse sources, including a prior, visual odometry (DVL), inertial measurement unit (IMU), depth sensors, and GPS.

Overcoming these limitations requires refined integrated sensor systems, mirroring the incremental strengthening of foundations that ensures a city’s resilience.

CoUGARs: A Scalable Architecture for Multi-Agent Autonomy

The CoUGARs (Collaborative Underwater Geophysical Robotic Systems) project addresses the need for scalable, multi-agent underwater robotic systems. It leverages a configurable, low-cost autonomous underwater vehicle (AUV) design, achieving a base cost of less than $3,000 USD per unit, enabling deployments of multiple units for coordinated data collection and complex task execution.

The system’s building blocks are individual AUVs, termed CougUVs, capable of independent operation, including autonomous navigation, sensor data acquisition, and onboard processing. These CougUVs are designed for collaborative tasks, communicating and coordinating actions beyond the capacity of a single robot.

A multi-agent field test was conducted at Spanish Oaks Reservoir, Spanish Fork, UT, in April 2025, utilizing two CougUVs, a modified BlueROV, and a base-station USBL system mounted on a PVC stand.
A multi-agent field test was conducted at Spanish Oaks Reservoir, Spanish Fork, UT, in April 2025, utilizing two CougUVs, a modified BlueROV, and a base-station USBL system mounted on a PVC stand.

The system’s software architecture is built upon ROS 2, providing a modular framework for communication, sensor integration, and algorithm development. This allows flexible configuration and integration of new capabilities, with distributed computing enabling collaborative data processing and decision-making.

Precision in Motion: Factor Graph-Based State Estimation

An autonomous underwater vehicle (AUV) requires robust state estimation for effective navigation and operation in complex environments. This is achieved through sensor fusion, combining data from inertial measurement units (IMUs), Doppler velocity logs (DVLs), and other sensors to create a more accurate and reliable representation of the vehicle’s position and orientation. A factor graph approach, implemented using the Georgia Tech Smoothing and Mapping (GTSAM) library, provides a statistically optimal estimate of the AUV’s state, handling sensor noise and environmental disturbances to estimate six degrees of freedom.

A CougUV successfully completed a waypoint mission in Utah Lake, Provo, UT, in November 2025, as demonstrated by its reported trajectory.
A CougUV successfully completed a waypoint mission in Utah Lake, Provo, UT, in November 2025, as demonstrated by its reported trajectory.

This architecture extends beyond localization, enabling simultaneous localization and mapping (SLAM) techniques to enhance navigational capabilities and situational awareness, providing valuable information for path planning, obstacle avoidance, and long-term autonomy.

Field Validation and Future Pathways in Underwater Robotics

The HoloOcean Simulator provides a comprehensive environment for testing and validating both the software and hardware components of the CoUGAR system prior to real-world deployment, allowing for robust pre-mission analysis and reducing operational risk. Successful operation has been demonstrated in over 20 field trials, with waypoint navigation and complex multi-agent coordination facilitated through integration with the MOOS-IvP suite.

During a mission in Bear Lake, Bear Lake, UT, in July 2025, the CougUV's depth and heading controllers exhibited responsive behavior.
During a mission in Bear Lake, Bear Lake, UT, in July 2025, the CougUV’s depth and heading controllers exhibited responsive behavior.

The current CoUGAR system achieves a theoretical depth rating of 60 meters. Future development efforts will focus on integrating acoustic communications capabilities, targeting a data transmission rate of 30 bytes every 4 seconds using Ultra-Short Baseline (USBL) technology, to further enhance multi-agent coordination. Additionally, exploration of 3D printing techniques is underway to enable rapid prototyping and customization of CougUV components, facilitating iterative design improvements.

Like the intricate network of vessels within a living body, the success of this system lies not in isolated improvements, but in understanding the flow between each component.

The CoUGARs platform, as detailed in the research, embodies a pragmatic approach to complex systems. It prioritizes functionality and configurability over sheer complexity, recognizing that a robust system isn’t built on eliminating constraints, but on understanding and managing them. This echoes G. H. Hardy’s sentiment: “A mathematician, like a painter or a poet, is a maker of patterns.” The CoUGARs platform isn’t simply a collection of hardware and software; it’s a carefully constructed pattern enabling research into multi-agent autonomy. The system’s modularity—allowing for adjustments to acoustic communication or SLAM algorithms—demonstrates that optimization in one area inevitably introduces tension elsewhere, necessitating a holistic understanding of the entire architecture to maintain overall stability and performance. It’s a living system, its behavior unfolding over time, far more significant than any static blueprint.

Where the Current Takes Us

The proliferation of low-cost robotic platforms, as exemplified by the CoUGARs system, presents a familiar paradox. Ease of access invariably shifts the focus from meticulous hardware design toward algorithmic complexity. While this accelerates innovation in areas like multi-agent localization and SLAM, it also risks obscuring fundamental limitations. Acoustic communication, particularly in challenging underwater environments, remains a bottleneck—a constraint not easily solved by software refinement. Future work must address the inherent trade-offs between platform cost, sensor fidelity, and communication bandwidth.

The current emphasis on factor graphs, while elegant, assumes a level of computational resources not always available in resource-constrained deployments. A move towards more parsimonious state estimation techniques, perhaps drawing inspiration from probabilistic robotics’ earlier days, may prove necessary. The challenge isn’t simply to map larger areas or track more agents, but to do so with a system that acknowledges its own limitations—a system that gracefully degrades rather than collapses under complexity.

Ultimately, the value of platforms like CoUGARs lies not in achieving perfect autonomy, but in providing a readily accessible testbed for exploring the boundaries of what is possible. The inevitable failures, the unexpected interactions, these are the true sources of insight. It is in understanding the fragility of these systems, rather than attempting to engineer them into invulnerability, that genuine progress will be made.

Original article: https://arxiv.org/pdf/2511.08822.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/

See also:

2025-11-13 22:21