⚡ electrical engineering

Tension-Aware Cooperative Trajectory Planning for Cable-Suspended Payload Transport by Two Underwater Vehicles

This paper presents a tension-aware cooperative trajectory planning framework for two six-degree-of-freedom underwater vehicles transporting a cable-suspended payload, which integrates full marine dynamics and a redundant differential-flatness parameterization to guarantee cable tautness through hard constraints in both kinodynamic search and spline optimization, thereby outperforming existing methods in cluttered, drag-dominated environments.

Original authors: Guoqing Huang, Yongfeng Li, Ning Li, Zhaobo Zhang, Wei Zhao, Jiuchao Zhang

Published 2026-07-31

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Guoqing Huang, Yongfeng Li, Ning Li, Zhaobo Zhang, Wei Zhao, Jiuchao Zhang

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the ocean floor not as a silent, empty void, but as a bustling construction site. Deep down, there are pipes to lay, cables to repair, and heavy machinery to move. Sometimes, the job is too big for a single robot. Just like two people might need to carry a heavy couch together, engineers want to use two underwater robots to haul a heavy load. But there's a catch: they can't just grab the load with rigid arms. Instead, they use cables. Think of these cables like the strings of a kite or the ropes on a tug-of-war team. They are light, simple, and strong, but they have one very strict rule: they can only pull. They can never push. If a cable goes loose, the whole operation turns into a chaotic, swinging mess, and the heavy load could get lost or damaged.

This is the tricky world of "cooperative manipulation" underwater. For robots flying in the air, scientists have figured out how to swing heavy loads with cables. But underwater is a different beast. The water is thick and heavy, acting like a giant, invisible hand that resists movement (drag) and changes how heavy things feel (buoyancy). If you try to use the same "air" rules underwater, the cables might go slack because the water pushes the load in ways the robots didn't expect. The big question for scientists is: How do you plan a path for two underwater robots so that they can carry a heavy object through a messy, obstacle-filled ocean without ever letting the cables go loose?

This paper, written by a team of researchers, answers that question with a clever new planning system. They created a "tension-aware" method that acts like a super-smart choreographer for two underwater robots. Instead of just telling the robots where to go, this system constantly calculates the invisible "pull" on the cables to make sure they stay tight and strong the entire time.

Here is how they did it and what they found. First, they realized that the old way of planning—where robots are treated like simple dots and cables are just straight lines—doesn't work underwater. The water's drag and the robots' own body movements change the game. So, the team built a new model that treats the robots as full, complex machines that feel the water's resistance.

The core of their solution is a mathematical trick called "redundant differential flatness." That's a mouthful, but think of it like this: usually, if you have two robots pulling a load, you might think the cables have to stay in a fixed shape. But the researchers realized the robots have a secret superpower: they can twist and turn their bodies to change the angle of the cables. They used this freedom to create a "safety map." This map ensures that no matter how the water pushes the load, the robots can always adjust their angles to keep the cables pulling hard enough to stay tight.

They tested this idea in a computer simulation that was like a video game for robots. They set up 50 different scenarios where the robots had to carry a heavy package through a forest of floating obstacles. They compared their new "tension-aware" planner against two other methods:

The "Fixed Cable" method: This is like trying to carry a couch with ropes tied to a rigid frame that can't move. In the simulation, this failed miserably. The cables went completely slack (dropping to a tension of -11.7 Newtons, which means they would have to push, which is impossible), and the load would have swung out of control.
The "Decoupled" method: This is like telling the robots where to go first, and then trying to figure out how to hold the ropes afterward. This kept the ropes tight, but the robots moved so fast and wildly that they crashed into the obstacles and spun their cables too fast (reaching 2.9 rad/s, which is nearly three times the safe limit).

The new planner, however, succeeded in 100% of the 50 test cases. It found paths where the robots moved smoothly, avoided all obstacles, and kept the cables pulling with a safe, positive tension (never dropping below 0.5 Newtons). The system worked by constantly adjusting the robots' positions and the cable angles together, rather than treating them separately.

To make sure this wasn't just a computer fantasy, the researchers also ran a "closed-loop" test. This is like putting the plan into a real robot controller and seeing if it works when things get messy. They added fake ocean currents and made the robots slightly heavier than they were supposed to be. Even with these surprises, the robots followed the path perfectly, and the cables stayed tight the whole time. The load drifted by less than a centimeter (0.8 cm) from the target path.

In short, this paper proves that by understanding the unique rules of the underwater world—specifically how water drag and robot movement affect cable tension—you can plan safe, smooth journeys for heavy loads. The key isn't just moving the robots; it's moving them in a way that respects the physics of the pull, ensuring the cables never go slack. While this was tested in simulations and not yet on real robots in the ocean, the results suggest that this method could soon help real underwater robots tackle the heavy lifting jobs of the deep sea.

Technical Summary: Tension-Aware Cooperative Trajectory Planning for Cable-Suspended Payload Transport by Two Underwater Vehicles

1. Problem Statement

The cooperative transport of heavy or bulky payloads via cables is a mature paradigm for aerial robots but remains an open challenge in the underwater domain. Existing underwater methods typically model robots as point masses and treat the unilateral nature of cables (pull-only) as a reactive check rather than a structural planning constraint. This approach fails in underwater environments due to three compounding factors:

Hydrodynamics: The presence of isotropic added mass, buoyancy, and nonlinear drag significantly alters the force balance required to move a payload.
Unilateral Constraints: Cables can only exert tension. If a planned trajectory drives tension to zero or negative, the cable goes slack, the payload becomes underactuated, and the system loses controllability.
Near-Neutral Buoyancy: Subsea payloads are often near-neutrally buoyant, meaning the required cable forces are small. Consequently, the margin to slack is thin, and standard planners often generate dynamically infeasible trajectories where cables would need to "push" the payload.

Furthermore, existing methods ignore the coupling between vehicle attitude and cable kinematics caused by body-fixed cable attachment offsets, and they fail to guarantee tension positivity during the planning phase.

2. Methodology

The authors propose a hierarchical, tension-aware planning framework for two full six-degree-of-freedom (6-DOF) Unmanned Underwater Vehicles (UUVs) transporting a point-mass payload. The framework integrates hydrodynamic dynamics, body-fixed offsets, and end-to-end tension guarantees.

System Modeling:

Dynamics: The system utilizes the Fossen vectorial model to incorporate isotropic added mass, restoring/buoyancy forces, and nonlinear drag.
Kinematics: Body-fixed cable offsets ( $r_{B,i}$ ) couple vehicle attitude to cable geometry. The system is modeled as a Differential-Algebraic Equation (DAE) where cables are massless rigid rods transmitting axial force.
Constraint: The taut-cable assumption imposes a holonomic distance constraint. The system must satisfy the unilateral condition $T_i \ge T_{min} > 0$ to remain in the fully-actuated mode.

Core Algorithm:
The framework consists of three main components:

Redundant Differential Flatness Parameterization:
- The authors define a redundant flat output $z = [p_L, \hat{l}_1, \hat{l}_2, \eta_1, \eta_2]$ , where $p_L$ is the payload position, $\hat{l}_i$ are cable unit vectors, and $\eta_i$ are vehicle attitudes.
- A single coplanarity constraint $(\hat{l}_1 \times \hat{l}_2)^T F_{req} = 0$ is introduced to close the one-dimensional redundancy. This ensures the required force $F_{req}$ lies in the plane spanned by the cables, allowing cable tensions to be recovered as a unique, closed-form quantity: $T = (L^T L)^{-1} L^T F_{req}$ .
- Proposition 1 establishes that while a fixed cable geometry is feasible only within a narrow cone of required forces, the freedom to re-orient the cables allows the system to maintain positive tension for any non-zero required force.
Tension-Aware Kinodynamic Front-End (Hybrid-A):*
- A search-based planner explores the flat-output space.
- The coplanarity constraint is resolved analytically to determine the out-of-plane payload jerk ( $j_{L,\perp}$ ), leaving 12 free control inputs.
- Pruning: Motion primitives are pruned if the recovered tensions violate $T_i \ge T_{min}$ or if collisions occur. This ensures only tension-feasible paths are expanded.
Constrained Spatiotemporal Back-End (Augmented Lagrangian):
- The front-end path serves as a warm start for a B-spline trajectory optimization.
- Coplanarity-Respecting Chart: The trajectory is parameterized using a chart where coplanarity and unit-norm constraints hold identically by construction (via Proposition 1). This reduces equality constraints to boundary conditions.
- Optimization: An Augmented Lagrangian Method (ALM) minimizes a cost function (payload jerk and cable angular acceleration) subject to hard inequality constraints: tension feasibility ( $T_i \ge T_{min}$ ), safe navigation corridors (SNCs) for collision avoidance, cable clearance, and kinodynamic bounds (velocity, acceleration, angular rates).

3. Key Contributions

First Full 6-DOF Underwater Model: The paper presents the first cooperative cable-transport planner that models carriers as full rigid bodies with body-fixed cable offsets and embeds hydrodynamic dynamics (added mass, drag) into the planning model.
Structural Tension Guarantee: Unlike prior work that treats tension feasibility as a reactive check, this method guarantees tension positivity end-to-end. It uses a redundant flatness parameterization and a specific geometric chart to make positive tension a structural property of the solution space.
Analytic Coplanarity Resolution: The method analytically resolves the cable-plane redundancy, converting the tension calculation into a closed-form expression and enabling efficient pruning in the front-end and hard-constraint enforcement in the back-end.
End-to-End Feasibility: The framework integrates a kinodynamic search with a constrained optimizer to produce trajectories that are collision-free, dynamically feasible, and strictly taut.

4. Results

The method was evaluated through numerical simulations, including a specific 3-D transport scenario, a buoyancy sweep, and a Monte Carlo study of 50 randomized tasks.

Baseline Failures:
- A fixed-cable baseline (minimum-jerk payload with fixed symmetric cables) failed on 100% of tasks. In the drag-dominated regime, it drove cable tensions to deep negative values (e.g., -11.7 N), implying the cables would need to push.
- A decoupled allocator (proposed path with post-hoc tension allocation) maintained positive tension but violated kinodynamic limits (cable angular rates up to 2.9 rad/s vs. 1.0 rad/s bound) and caused vehicle-obstacle collisions because it did not co-design the cable orientation with the path.
Proposed Method Performance:
- Feasibility: The proposed planner achieved 100% feasibility on all 50 randomized tasks, satisfying tension margins, coplanarity, collision avoidance, and kinodynamic bounds.
- Tension Maintenance: In the primary scenario, the minimum tension was maintained at 0.69 N (above the 0.5 N threshold) throughout the transport.
- Robustness: A buoyancy sweep showed the method maintains tension margins down to near-neutral buoyancy ( $W_{net} \approx 1$ N), whereas other methods failed or collapsed to zero tension.
- Closed-Loop Tracking: A geometric SE(3) controller tracked the generated trajectory in a full multibody simulation with model mismatch and current disturbances. The payload tracking error was 0.8 cm, and measured cable tensions remained strictly positive (minimum 0.58 N), confirming the cables stayed taut.

5. Significance and Claims

The paper claims to be the first to address cooperative cable transport for underwater vehicles with a model that includes full 6-DOF dynamics, body-fixed offsets, and an end-to-end guarantee of cable-tension positivity.

The authors assert that their work demonstrates that unilateral feasibility is a genuine obstacle for near-neutral underwater payloads in drag-dominated regimes, which defeats fixed-geometry approaches. They further claim that simply reorienting cables post-hoc is insufficient; the orientation must be co-designed with the payload path to respect rate and collision limits.

The significance of the work lies in its ability to generate, prior to execution, a collision-free transport trajectory that, by construction, keeps both cables above a user-specified minimum tension despite buoyancy, drag, and vehicle acceleration limits. The method outputs reference trajectories and feedforward control inputs that integrate directly with standard SE(3) vehicle controllers, offering a practical solution for subsea tasks where loads are too large for a single vehicle.

The authors note that the current implementation is an offline planner with solve times on the order of tens of seconds, and future work will focus on compiled real-time implementations, extension to $n > 2$ vehicles, and robustness against unmodeled currents and cable elasticity.