Sailing forward: A review of contrasts and synergies between racing and robotic sailing

Wed, 01 Apr 2026 12:02:00 +0800

For thousands of years, humanity’s use of wind energy for navigation has run through the entire Age of Sail. However, driven by modern technology, the ancient sport of sailing has quietly diverged into two parallel development trajectories in scientific research: one is racing sailing, which pursues ultimate speed and human limits; the other is autonomous sailboats/unmanned sailboats, which target unknown waters and seek long-term autonomous operations.

Figure 1. (Left) Racing sailboat, (Right) Unmanned sailboat

Both rely on wind propulsion and follow the basic laws of fluid dynamics and aerodynamics, yet engineers and scientists in these two fields often work in isolation. Due to significant differences in research objectives and application scenarios, both sides have established relatively closed academic circles, leading to a methodological divide. The racing field has accumulated massive amounts of high-fidelity aerodynamic/hydrodynamic data and expert experience, while the unmanned sailboat field has developed unique expertise in autonomous control and intelligent planning algorithms.

If this disciplinary barrier is broken, can the two achieve technological crossover and complementarity? This study conducts the first comprehensive cross-domain comparison between racing sailboats and unmanned sailboats, systematically sorting out the technical differences, integration potential, and common challenges faced by both.

 1. **Different Paths to the Same Goal: Distinctly Different Design Philosophies and Application Scenarios**

To understand the gap between these two fields, it is first necessary to analyze the differences in their underlying logic:

l Racing sailboats (ocean racing yachts and inshore dinghies): The core constraint is the race rules, and the core goal is the most efficient energy conversion and speed limits. Guided by this, researchers spare no expense in conducting wind tunnel tests, towing tank tests, and supercomputer computational fluid dynamics (CFD) simulations. Their system operation heavily relies on the keen intuition and extreme micro-maneuvering of top human sailors under complex sea conditions.

l Unmanned sailboats: The core constraints are energy and harsh sea conditions, and the core goals are long-term survival and mission reliability. Under unmanned conditions, platforms often adopt conservative designs (e.g., deep keels with bulbs, rigid wing sails). The main research force consists of experts in robotics, automation control, and marine engineering, who focus more on robust control algorithms and obstacle avoidance navigation.

This circle isolation leads to a significant misallocation of resources. Many optimization strategies that are mature in the racing field are still in the trial-and-error stage in the unmanned sailboat field; conversely, advanced perception and decision-making algorithms from the robotics field are rarely fed back into traditional sailing training.

 2. **Collision of Four Technical Dimensions and Potential for Cross-Domain Transfer**

To systematically explore the pathways for integration between the two, the research team proposed a hierarchical analysis framework, deeply analyzing the methodological differences between the two from four core technical dimensions and assessing the feasibility and main obstacles of cross-domain technology transfer: 2.1 Component Performance Evaluation

In the performance evaluation of sails, hulls, and appendages (such as keels and foils), racing sailboats, backed by deep capital and racing experience, dominate high-precision experimental and simulation workflows. In contrast, unmanned sailboats, constrained by development cycles and budgets, often rely on simplified two-dimensional airfoil data or empirical formulas.

v Cross-domain Insights: The mature CFD simulation pipelines and laboratory evaluation methods from the racing field have high downward compatibility and can be readily applied to the development of unmanned sailboats, reducing their early-stage trial-and-error costs.

Figure 2. The nonlinear system-level coupling between propulsion, resistance, and heading leads to implicit performance that cannot be directly predicted by analytical methods. 2.2 Overall Performance Evaluation

In the nonlinear coupled marine environment, how can the overall motion performance of a vessel be predicted? The racing field widely uses Velocity Prediction Programs (VPP) and their dynamic versions (DVPP), mapping the relationship between wind speed, wind direction, and optimal boat speed through vast databases.

v Cross-domain Insights: Due to the diverse platform forms and extreme operating sea conditions of unmanned sailboats, prior data is lacking. Directly applying traditional VPP faces the curse of dimensionality. In the future, developing variable-fidelity dynamic models that balance computational efficiency and accuracy by combining machine learning and adaptive sampling techniques is a common R&D goal for both fields.

Figure 3. The speed of a sailboat varies greatly under different wind speeds and headings, necessitating comprehensive, full-envelope performance evaluation.

(Figure note: Sailing faces strong nonlinear coupling of aerodynamic forces, hydrodynamic forces, and attitude, requiring a precise system-level performance evaluation framework for support.) 2.3 Motion Control

This is where the differences between the two are most pronounced. Racing sailboats rely entirely on the model-free intuitive control of human experts; sailors execute complex tactical maneuvers through delicate heel and sail trimming. In contrast, unmanned sailboats depend on rigid sensor feedback and automatic control algorithms like PID and Model Predictive Control (MPC), with strategies usually being extremely conservative to prevent capsizing.

v Cross-domain Insights: The research team proposed a highly forward-looking vision: using imitation learning or deep reinforcement learning to enable unmanned sailboats to learn the multi-degree-of-freedom cooperative operation strategies of top human sailors. This could not only unlock the sailing potential of unmanned sailboats but also, conversely, create tireless, high-level AI virtual sparring partners, like AlphaGo, for professional sailors, advancing the evolution of human sailing technology.

Figure 4. Typical mechanical control systems in sailboats: (a) Wind vane self-steering system: passively maintains heading via a wind vane mechanically linked to the rudder; (b) Wave-piercing foils: provide roll stability using lift changes related to immersion depth; (c) Height adjustment system: mechanically adjusts sailing draft via a feeler linked to the foil flap; (d) Adaptive wing sail: achieves passive wind adjustment using aerodynamic feedback between the main wing and tail wing. Advanced Cooperative Control Techniques in Racing Sailboats

While the algorithms of unmanned sailboats are still struggling to calculate how not to capsize, human sailors can already use the edge of capsizing to accelerate and use the opposition of wind and rudder to brake. This model-free control based on physical intuition is precisely the huge gap that we point out in the paper that the control systems of unmanned sailboats need to bridge in the future.

1. Roll-tacking: A cornering technique that uses heel to gain acceleration, shown below (left)

Sailboats have no throttle, and every turn usually means deceleration. But when top sailors perform a sharp tack upwind, they don’t rigidly slam the rudder over. Instead, they deliberately induce a sharp heel first, using the asymmetric hull shape to naturally turn the bow into the new course (like leaning a bicycle into a turn, avoiding the drag from rudder movement). At the precise moment the bow turns, the sailor leaps forcefully to the other side, which is rising high, using their full body weight to instantly flatten the boat. This violent flattening action is like swinging a giant fan, causing the sail to whip through the air, artificially creating an extra gust of wind that slingshots the boat out of the turn, accelerating rather than decelerating.

2. Heaving-to: A maritime brake using the opposition of wind and water, shown below (right)

Sailboats have no brake pedal. What to do if you need to stop to rest or make emergency repairs in a gale? Sailors cleverly use natural forces to lock the boat: they deliberately pull the headsail to the wrong side, letting the wind try to push the bow backward; simultaneously, they jam the underwater rudder hard in the opposite direction, letting the water try to push the bow forward. These two opposing forces perfectly cancel out on the hull. The boat not only loses forward momentum but also sits steadily, hove-to at a safe angle. This slow drifting motion creates a slick of calm water on the windward side, providing a natural haven for the crew amidst the storm.

Figure 5. Expert sailors can perform complex maneuvers through exquisite body weight distribution and coordination of sail and rudder. Translating this human experience into AI algorithms is currently a blue ocean in the control field. 2.4 Reactive Navigation and Weather Routing

In trajectory planning, racing sailing is essentially a game, requiring the integration of high-resolution local weather models and the use of rules like wind shadow effects and right-of-way to suppress opponents. Unmanned sailboats, on the other hand, mostly present as single-vessel optimal path planning, focusing on static/dynamic obstacle avoidance and utilizing ocean currents to save energy. Despite the different logic, both fields share highly consistent technical pain points when facing the challenge of online re-planning under dynamic wind fields and severe sea states.

Figure 6. Sailboat route planning: (Left) Upwind zigzag maneuver; (Right) Wind shadow effect of nearby vessels.

Common Future Challenges and Implications for Zero-Carbon Shipping

Based on the systematic comparison above, the research team outlined a clear cooperation roadmap for future wind propulsion technology, pointing out that both sides should conduct joint research in the following three frontier directions:

(1) Build a unified modeling framework that balances accuracy and computational cost;

(2) Develop dynamic models capable of adapting to complex wave disturbances and near-field situational awareness technologies;

(3) Accelerate the deep integration of human maritime expert experience with embodied intelligence.

The strategic value of this cross-domain research extends far beyond sailboat racing and ocean observation. Against the backdrop of the global pursuit of “dual carbon” goals, modern ocean-going commercial vessels are undergoing a green revolution in wind-assisted propulsion. Whether it’s giant rotor sails or rigid wing sails, the core aerodynamic mechanisms, system coupling modeling, and multi-objective weather routing algorithms are precisely the areas that racing sailboats and unmanned sailboats have been deeply cultivating for years. The technological convergence of these two major research camps will undoubtedly lay a crucial technical cornerstone for the next generation of sustainable intelligent shipping.

This review paper, titled “Sailing forward: A review of contrasts and synergies between racing and robotic sailing,” was published in the top journal in the field of ocean engineering, the Journal of Ocean Engineering and Science. The first author of the paper is postdoctoral fellow Yang An (currently working at the Institute of Deep-sea Science and Engineering, Chinese Academy of Sciences), and the corresponding author is Associate Professor Zhengru Ren.

Article link: https://doi.org/10.1016/j.joes.2026.03.008

Autonomous Sailboat | Zhengru Ren

Sailing forward: A review of contrasts and synergies between racing and robotic sailing