A team of engineers and biologists has used the Stampede3 supercomputer at the Texas Advanced Computing Center (TACC) to uncover how hummingbirds perform lightning‑fast escape maneuvers, dissecting the balance between passive flapping counter torque and active neuromuscular control in some of the most agile fliers in nature.
In their new Bioinspiration and Biomimetics article, Flapping counter torque and active control in the escape maneuvers of hummingbirds, researchers from Vanderbilt University, the University of Montana and Pennsylvania State University reconstructed real escape flights from high‑speed video and then subjected those maneuvers to intensive computational fluid dynamics (CFD) analysis. The team used U.S. National Science Foundation (NSF) ACCESS allocations on Stampede3 to simulate both “free‑body” flight, in which the bird’s body is allowed to rotate, and “fixed‑body” flight, in which body‑rotation‑induced wing velocities are numerically removed while preserving the wing motion relative to the body.
“Thanks to the NSF ACCESS allocations, we were able to perform and compare these two sets of simulations,” said Haoxiang Luo, professor and chair of Vanderbilt University’s Department of Mechanical Engineering. “We separated passive flapping counter torque (FCT) from actively generated aerodynamic torques and our results show that FCT in the roll axis is particularly strong during downstroke because body rotation produces large left‑right wing velocity asymmetries and alters wing angles of attack in multiple axes.”
Ultimately, we are showing how to turn high‑speed animal motion into actionable engineering principles for the next generation of agile autonomous systems.
–Haoxiang Luo, Vanderbilt University
Luo explained that to maintain and steer rapid rotations despite this strong damping, hummingbirds generate active torques that counteract FCT during downstroke and adopt wing kinematics that reduce FCT during upstroke, enabling both stability and extreme agility in the same maneuver.
How TACC’s Stampede3 Enabled the Team’s Work
Capturing these effects required high‑fidelity CFD simulations at fine spatial and temporal resolution over multiple wingbeats and various wing-body configurations, quickly pushing beyond the capacity of desktop hardware. Each scenario couples detailed 3D wing and body kinematics with unsteady aerodynamics that involves extensive flow separation and vortex shedding from wing surfaces.
Stampede3’s CPU- and GPU-accelerated architecture provided the throughput needed to efficiently execute large model ensembles. With more than 1,800 compute nodes and over 140,000 cores, the system is designed to support the kind of data- and compute-intensive CFD workloads at the heart of this study.
Inside the Workload
On Stampede3, the research team distributed escape maneuvers – combining rapid pitch, roll, yaw and linear accelerations – across many nodes, transforming what would have been a months-long serial effort into an efficient parallel workflow. The system’s mix of high-bandwidth memory CPU nodes allowed the group to select the optimal hardware for each stage, from core CFD solves to post-processing aerodynamic force and torque histories. These capabilities enabled both fixed-body and free-body simulations at the spatial and temporal resolution needed to resolve subtle differences between passive and active torques.
From Animal Flight to Agile Machines
By quantifying how much of a hummingbird’s escape response comes from passive flapping counter torque versus active control, and how that balance shifts across axes and phases of the wingbeat, the work offers concrete design targets for bioinspired flapping‑wing robots and micro air vehicles.
“Designers can use these insights to shape wings and body dynamics that supply ‘built‑in’ rotational damping for disturbance rejection, while reserving active control for only those aspects of the maneuver that truly demand fast actuation and sensing,” Luo said. “Ultimately, we are showing how to turn high‑speed animal motion into actionable engineering principles for the next generation of agile autonomous systems.”
Resource Provider Institution(s): Texas Advanced Computing Center (TACC)
Resources Used: Stampede3
Affiliations: Vanderbilt University, University of Montana, Pennsylvania State University
Funding Agency: NSF
Grant or Allocation Number(s): PHY230028 and CTS110025
The science story featured here was enabled by the U.S. National Science Foundation’s ACCESS program, which is supported by National Science Foundation grants #2138259, #2138286, #2138307, #2137603, and #2138296.