Airflow-mediated robotic manipulation
main / research / airflow-field manipulation
Manipulating objects remotely using airflow has long appeared in fiction, but systematic robotic manipulation using airflow fields remains relatively unexplored. Existing demonstrations have often focused on specific aerodynamic phenomena, such as the Coandă effect, while airflow itself is already widely used in industry and everyday life for tasks such as cleaning, dust removal, and leaf collection.
In our research, we address the challenge of turning airflow into a controllable medium for robotic manipulation. We have developed wind-mimicking robotic technologies that use directed airflow fields to manipulate objects remotely over distances of up to several meters, across a wide range of object sizes, shapes, and materials.
Our initial approach uses visual feedback to steer a jet-induced airflow field according to the observed position and motion of the object. By controlling the orientation of an air nozzle relative to the object, the airflow can drive the object in a desired direction, including toward the nozzle rather than simply blowing it away. This enables objects to follow different trajectories, including circular paths and complex letter-like patterns. The approach is highly versatile: we have manipulated regularly shaped polystyrene hemispheres and sticks, irregular objects such as cotton wads and face masks, and deformable objects including crumpled tissue paper and plastic bags.
We have subsequently developed model-based methods to extend the controllability and scalability of airflow-mediated manipulation. Instead of relying only on the instantaneous object response, these methods combine a model of the jet-induced airflow field with learned object dynamics and model-based predictive control. This makes it possible to predict how objects respond to changes in the airflow and to control multiple objects simultaneously at meter-scale distances.
This form of airflow-mediated robotic manipulation is particularly attractive when direct physical contact is difficult, undesirable, or impossible. Beyond trajectory-following on solid surfaces, we have demonstrated manipulation on water surfaces and across solid-water environments, as well as manipulation in the presence of disturbing airflows. We have also demonstrated more complex automated tasks, including collecting objects into a target receptacle, steering a tethered mobile agent to hook and retrieve heavier objects, and maneuvering a tethered soft agent to close an electrical circuit.
Together, these results show that airflow can be treated not merely as a disturbance or a simple blowing mechanism, but as a controllable physical medium through which robots can interact with objects remotely.
Selected publications:
- Haeri, S., Kopitca, A., Kandemir, H. and Zhou, Q., "Meter‐Scale Distance Manipulation of Diverse Objects with Jet‐Induced Airflow Field", Advanced Intelligent Systems, p.2400174, 2024.
- Kopitca, A., Haeri, S., and Zhou, Q., "Remote Manipulation of Multiple Objects with Airflow Field Using Model-Based Learning Control", IEEE/ASME Transactions on Mechatronics, vol. 30, no. 4, pp. 2871 - 2879, 2025.
- Kopitca, A., Haeri, S., Iqbal, S., Zang, X., Zeng, H., Zhou, Q., "3D Manipulation of Airborne Objects Using Patterned Airflow", International Conference on Manipulation, Automation and Robotics at Small Scales, MARSS 2026.