Wake Up to the Past: Using Memory to Model Fluid Wake Effects on Robots

Luca Vendruscolo; Eduardo Sebastián; Amanda Prorok; Ajay Shankar

cs.ROMar 23, 2026

Wake Up to the Past: Using Memory to Model Fluid Wake Effects on Robots

Authors: Luca Vendruscolo, Eduardo Sebastián, Amanda Prorok, Ajay Shankar

Organizations: Department of Computer Science and Technology, University of Cambridge, UK.

Abstract

Autonomous aerial and aquatic robots that attain mobility by perturbing their medium, such as multicopters and torpedoes, produce wake effects that act as disturbances for adjacent robots. Wake effects are hard to model and predict due to the chaotic spatio-temporal dynamics of the fluid, entangled with the physical geometry of the robots and their complex motion patterns. Data-driven approaches using neural networks typically learn a memory-less function that maps the current states of the two robots to a force observed by the "sufferer" robot. Such models often perform poorly in agile scenarios: since the wake effect has a finite propagation time, the disturbance observed by a sufferer robot is some function of relative states in the past. In this work, we present an empirical study of the properties a wake-effect predictor must satisfy to accurately model the interactions between two robots mediated by a fluid. We explore seven data-driven models designed to capture the spatio-temporal evolution of fluid wake effects in four different media. This allows us to introspect the models and analyze the reasons why certain features enable improved accuracy in prediction across predictors and fluids. As experimental validation, we develop a planar rectilinear gantry for two spinning monocopters to test in real-world data with feedback control. The conclusion is that support of history of previous states as input and transport delay prediction substantially helps to learn an accurate wake-effect predictor.

Figures & tables

Panel of Table 1 — TABLE I : Summary of model architectures. Parameter counts refer to trainable weights.

Fig. 2 : Performance comparison of each of the methods on the four domains. ( left ) A CFD dataset for two quadrotors [ 9 ] . ( center left ) A numerical model of ship-ship interactions [ 37 ] . ( center right ) A dataset that captures hydrodynamic interactions between two fish swimming in line [ 38 ] . ( right ) Closed-loop tracking on a custom 2D downwash simulator.

Panel of Figure 3 — Fig. 3 : Fish schooling analysis. ( left ) Scatter plot of true vs. predicted disturbances. The memoryless MLP ( $R^{2}=0.57$ ) exhibits two distinct modes, failing to distinguish periodic phases, while the GRU achieves accurate, unbiased predictions. ( right ) Convergence basin of the Delay Embedding model. Depending on the random initialisation of $\mu$ , the model converges to either the true physical delay (0.33 s, higher $R^{2}$ ) or an alias peak (0.148 s, lower $R^{2}$ ).

Panel of Figure 4 — Fig. 4 : ( left ) Average attention weights of the Cross-Attention model over all episodes in the 2D downwash simulator. The profile shows a clear peak near the true physical delay (red dashed line at $0.38\text{\,}\mathrm{s}$ ), demonstrating that the model attends to the correct time lag. ( right ) Learned versus physical delay for the Delay Embedding model across five values of $\Delta z$ .

Fig. 5 : Out-of-distribution closed-loop ablation on the 2D downwash simulator. Models are trained with physics parameters sampled at

\pm 10\%

of their nominal values and evaluated at

\pm 10\%

\pm 50\%

, and

\pm 75\%

perturbation levels. Red lines denote the baseline closed-loop RMSE from a 3-seed evaluation trained and evaluated at the nominal (unperturbed) conditions. Degradation preserves the relative ranking across conditions.

Panel of Figure 6 — Fig. 6 : The tethered monocopters gantry system. (top left) The wake source can move freely in the X axis, while the sufferer can move freely in the XZ plane. The sensors interface with an Arduino UNO and a Raspberry Pi 5 to implement feedback control with/without disturbance rejection. (top right) Four overlay snapshots from an episode where the sufferer attempts to track a reference trajectory (dashed red) and deviates significantly (green curve) due to the time-varying turbulence caused by the wake source. (bottom left) Memory-based predictors achieve better open-loop prediction performance than the memory-less baseline, matching numerical and simulated evaluations in Figure 2 . (bottom right) Mean altitude trajectory across $n{=}5$ repeated PID-only episodes on the same randomization seed. The dashed red line indicates the target altitude; the solid green curve shows the mean position; the shaded region denotes $\pm 1\sigma$ . The characteristic downwash-induced dip is clearly visible.

Multi-Robot Systems

Multi-robot systems coordinate teams of robots to plan, perceive, communicate, and act together in shared tasks and environments.

See all 205 papers

Real Time Local Wind Inference for Robust Autonomous Navigation
Apr 1, 2026Spencer FolkAerial Robotics Robot Navigation
Mechanical Engineering and Applied Mechanics, University of Pennsylvania
The balance between compactness and forecast accuracy of data-driven latent-space reduced-order models in controlled wake flows
Jul 27, 2026Alberto Solera-Rico, Patricia García-Caspueñas, Carlos Sanmiguel Vila +1Variational Autoencoders Latent Dynamics Modeling
Sub-directorate general of aeronautical systems, Spanish National Institute for Aerospace Technology (INTA), ctra. M-301, km. 10.5, 28330 San Mart´ın de la Vega, Madrid, Spain. · Departamento de Ingenier´ıa Aeroespacial, Universidad Carlos III de Madrid, Avenida de la Universidad 30, 28911 Legan´es, Madrid, Spain. · Department of Mechanical Engineering, University of Washington, 1410 NE Campus Parkway, Seattle, WA 98195, USA.
FlowMo-WM: A World Model with Object Momentum and Hidden Ambient Drift
Jun 11, 2026Yitao Jiang, Luyang Zhao, Muhao Chen +1World Model Learning World Models for Robotics
Department of Computer Science Dartmouth College, Hanover, NH · Department of Electrical and Computer Engineering Clemson University, Clemson, SC · Department of Mechanical and Aerospace Engineering University of Houston, Houston, TX

Apr 1, 2026J/K move · Enter open · S save

Real Time Local Wind Inference for Robust Autonomous Navigation

Spencer Folk

Mechanical Engineering and Applied Mechanics, University of Pennsylvania

Aerial Robotics Robot Navigation Spatiotemporal Forecasting

Open paper AskPDF arXiv

This thesis presents a solution that enables aerial robots to reason about surrounding wind flow fields in real time using on board sensors and embedded flight hardware. The core novelty of this research is the fusion of range measurements with sparse in situ wind measurements to predict surrounding flow fields. We aim to address two fundamental questions: first, the sufficiency of topographical data for accurate wind prediction in dense urban environments; and second, the utility of learned wind models for motion planning with an emphasis on energy efficiency and obstacle avoidance. Drawing on tools from deep learning, fluid mechanics, and optimal control, we establish a framework for local wind prediction using navigational LiDAR, and then incorporate local wind model priors into a receding-horizon optimal controller to study how local wind knowledge affects energy use and robustness during autonomous navigation. Through simulated demonstrations in diverse urban wind scenarios we evaluate the predictive capabilities of the wind predictor, and quantify improvements to autonomous urban navigation in terms of crash rates and energy consumption when local wind information is integrated into the motion planning. Sub-scale free flight experiments in an open-air wind tunnel demonstrate that these algorithms can run in real time on an embedded flight computer with sufficient bandwidth for stable control of a small aerial robot. Philosophically, this thesis contributes a new paradigm for localized wind inference and motion planning in unknown windy environments. By enabling robots to rapidly assess local wind conditions without prior environmental knowledge, this research accelerates the introduction of aerial robots into increasingly challenging environments.

Abstract

Multi-Robot Systems

Explore similar work

Real Time Local Wind Inference for Robust Autonomous Navigation

The balance between compactness and forecast accuracy of data-driven latent-space reduced-order models in controlled wake flows

FlowMo-WM: A World Model with Object Momentum and Hidden Ambient Drift