Efficient wave manipulation with sparse control under partial observability

Recent advancements in control and machine learning have accelerated progress in wave manipulations. These advancements include the use of deep neural networks to represent dynamics that are partially observed by robot sensors, as well as effective control using sparse control signals. In this talk, we will present our research on the manipulation of time-varying acoustic waves, which are partially observed by a robot capable of influencing the waves through spatially sparse actuators. This problem holds great potential for new material designs, ultrasonic imaging, and other applications. We develop an efficient data-driven method for robot learning that applies to either focusing scattered energy in a designated region or suppressing it [1]. This method is better in terms of solution quality and computational complexity as compared to a state-of-the-art learning based method for the manipulation of dynamical systems. We compare the control capabilities of our physics-informed approach with a well-known physics-uninformed baseline and show that our method is more robust in a variety of scenarios. The project codes are available at: https://gladisor.github.io/waves/. [1] Shah, T., Smilovich, N., Amirkulova, F., Gerges, S., Tiomkin, S. Acoustic Wave Manipulation Through Sparse Robotic Actuation. IEEE ICRA2025, arXiv preprint arXiv:2502.08784, 2025.