⚡ electrical engineering

Speed-Weighted Adaptive Flocking for Sailing Swarms under Dynamic Environmental Forcing

This paper introduces the SailSwarmSwIM simulator and a novel speed-weighted flocking algorithm to address the coordination challenges of autonomous sailing robots, demonstrating that prioritizing the social influence of slower robots effectively maintains flock cohesion and polarization under dynamic wind conditions.

Original authors: Pranav Kedia, Aaron Gan, Hannah J. Williams, Andreagiovanni Reina, Heiko Hamann

Published 2026-05-28

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Pranav Kedia, Aaron Gan, Hannah J. Williams, Andreagiovanni Reina, Heiko Hamann

Original paper licensed under CC BY 4.0 (http://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 a flock of birds trying to fly together. In most computer models, we assume every bird is identical: they all fly at the same speed, turn instantly, and can go in any direction they want. But the researchers in this paper are studying something very different: a flock of robotic sailboats.

Here is the problem: Sailboats are not like birds. They are at the mercy of the wind.

The Wind is the Boss: If the wind is weak, a sailboat turns very slowly. If the wind is strong, it turns quickly.
The "No-Go" Zone: A sailboat cannot sail directly into the wind. If it tries, it gets stuck. It has to zigzag (a move called "tacking") to make progress.
The Chaos: In a group of boats, one boat might be in a gust of wind and zooming ahead, while its neighbor is in a calm patch and barely moving. If they all try to follow the same rules, the fast ones will pull away, and the slow ones will get left behind, causing the flock to break apart.

The Solution: A New Way to "Listen" to Each Other

The researchers created a new rule for how these robots should talk to each other. They call it "Speed-Weighted Adaptive Flocking."

Think of it like a group of friends hiking up a mountain.

The Old Way (The Baseline): Everyone listens to the group equally. If the fast hikers keep walking at their own pace, the slow hikers get left behind. The group stretches out and eventually breaks up.
The New Way (Speed-Weighted): The group decides to pay extra attention to the slowest hikers. The fast hikers don't just ignore the slow ones; they actively slow down and wait for them.

In the paper's model, they introduced a "knob" (called $\gamma$ ) that controls how much weight the robots give to their neighbors' speed.

If you turn the knob to listen to fast neighbors, the group stretches out and becomes disorganized.
If you turn the knob to listen to slow neighbors, the group stays tight and safe, but they might get a little confused about which direction to go.
The Sweet Spot: The researchers found a "Goldilocks" setting (a very small positive number) where the robots listen slightly more to the slow ones. This keeps the group together (cohesion) and prevents crashes (safety) without making them lose their direction (polarization).

The Simulator: "SailSwarmSwIM"

Before testing this on real boats (which is expensive and risky if the weather is bad), the team built a video game-like simulator called SailSwarmSwIM.

It's a "reduced-order" model, meaning it's a simplified version of reality that focuses only on the most important parts: wind speed, wind direction, and the fact that boats can't sail straight into the wind.
It simulates different weather scenarios: steady breezes, strong winds, and gusty, unpredictable winds.

What They Found

They ran thousands of simulations with 10 robots in different wind conditions. Here is what happened:

The "Fast Neighbor" Trap: When the robots tried to follow the fastest boat, the flock stretched out like a rubber band until it snapped. The group became disorganized.
The "Slow Neighbor" Anchor: When they weighted the influence toward the slowest boats, the flock stayed compact.
The Perfect Balance: By slightly favoring the slow neighbors, they achieved the best of both worlds:
- Safety: The boats crashed into each other far less often.
- Cohesion: The group stayed together instead of spreading out.
- Alignment: The group still managed to move in the same general direction.

The "Sail Luffing" Trick

There was one extra trick they used. Even if the robots agreed on a direction, a boat with a perfect wind angle might naturally go faster than a boat struggling against the wind. To fix this, the robots have a mechanism called "sail luffing."

Imagine a sailboat "loosening" its sails to catch less wind and slow down. The researchers programmed the fast boats to automatically "loosen their sails" to match the speed of the slower neighbors. This ensures that even if they all want to go the same way, they actually move at the same speed.

The Bottom Line

The paper concludes that for a swarm of wind-powered robots, the best strategy isn't to let the fastest member lead the charge. Instead, the group should be anchored by its slowest, most constrained members. By giving a little extra weight to the slow neighbors, the whole flock becomes safer, stays together better, and moves more effectively through changing wind conditions.

They tested this in their simulator and found that a specific, moderate setting works best across all types of weather, from calm breezes to strong gusts.

Technical Summary: Speed-Weighted Adaptive Flocking for Sailing

Problem Statement
Classical collective motion models (e.g., Reynolds, Vicsek, Couzin) typically assume self-propelled agents with homogeneous locomotion capabilities that can directly execute desired speeds and headings. This assumption fails for autonomous sailing robots, whose motion is fundamentally constrained by environmental forcing. Specifically, sailing robots face:

Wind-dependent propulsion: Speed and maneuverability are coupled to wind speed; weak winds result in slow or impossible turns, while strong winds enable faster maneuvers.
Restricted headings: Robots cannot sail directly upwind (the "no-go" zone) and must perform tacking maneuvers.
Transient heterogeneity: Even in a physically homogeneous fleet, spatially and temporally varying wind conditions create temporary differences in speed and maneuverability across the swarm.

This creates a "fast-slow coordination" problem: faster robots risk stretching or fragmenting the flock, while slower, constrained robots risk being left behind. The core challenge is determining how social influence should be weighted—should the flock follow the fastest neighbors to maintain progress, or anchor around the slowest to maintain cohesion?

Methodology
To address this, the authors developed SailSwarmSwIM, a reduced-order simulator designed specifically for autonomous sailing robot swarms. The simulator captures wind-dependent speed, no-go zones, tacking behavior, and both steady and gusty wind fields without modeling full six-degree-of-freedom hydrodynamics.

The study compares a baseline Couzin-type flocking controller against a novel Speed-Weighted Adaptive Flocking approach:

Baseline: Uses standard zonal interactions (repulsion, orientation, attraction) with uniform social weighting. Speed variations arise passively from wind dynamics.
Speed-Weighted Controller: Introduces a social interaction rule where a neighbor's influence is weighted by their current speed ( $v_j$ $v_{j}$ ) using a regularized inverse-power law: $w_{ij} = 1/(v_j + \epsilon)^\gamma$ $w_{ij} = 1/ (v_{j} + ϵ)^{γ}$ .
- $\gamma < 0$ : Prioritizes faster neighbors (fast-neighbor following).
- $\gamma = 0$ : Uniform weighting (recovered Couzin baseline, though with added luffing).
- $\gamma > 0$ : Prioritizes slower neighbors (slow-neighbor anchoring).
Speed Equalization: The controllers include a "sail luffing" layer that actively reduces propulsion for faster boats (by easing the sail) to match neighborhood speeds, reducing heading-induced speed dispersion.

Experiments were conducted with $N=10$ robots across four wind conditions (5 m/s and 10 m/s, both steady and gusty). The exponent $\gamma$ was swept from $-2$ to $10$. Performance was evaluated using three metrics: flock compactness (convex hull area), alignment (polarization), and safety (cumulative unsafe proximity events).

Key Results
The study identified a "robust compromise" regime where moderate slow-neighbor anchoring improves overall flock performance:

Optimal Range: A narrow band of small positive exponents ( $\gamma \in [-0.1, 0.3]$ ) yielded significant improvements across all metrics simultaneously.
Performance Gains: At the representative setting $\gamma = 0.01$ $γ = 0.01$ , the flock achieved:
- Compactness: A 15–29% reduction in convex hull area compared to the baseline.
- Safety: A 22–32% reduction in unsafe proximity events.
- Alignment: A modest but significant increase in polarization.
Failure Modes:
- Strong Fast-Neighbor Following ( $\gamma \le -1.5$ ): Caused significant flock stretching and reduced polarization, particularly in high-wind environments.
- Strong Slow-Neighbor Anchoring ( $\gamma \ge 1$ ): Preserved cohesion but progressively eroded alignment (polarization), leading to a disordered group.
Mechanism: The improvement arises from a balance where weighting slower neighbors pulls the flock's centroid toward constrained robots (preventing fragmentation) while reducing the relative speed between leaders and trailers (lowering collision risks).

Significance and Claims
The paper claims to provide a modeling framework for studying adaptive collective behavior in robotic fleets where motion capabilities are continuously shaped by the environment. The primary contribution is demonstrating that increasing the social influence of slower robots improves polarization and reduces close encounters in wind-driven swarms.

The authors emphasize that this approach does not seek a universally optimal controller for every single metric but rather identifies a regime that balances cohesion, safety, and alignment. The work validates that accounting for transient locomotor heterogeneity through speed-weighted social rules is essential for maintaining coherent group behavior in sailing robot swarms, a problem where classical homogeneous models are insufficient. The findings are derived from simulation, with the authors noting that future work will address heterogeneous fleets, energy costs, and physical field experiments.