arXiv subjects

Michael A. Bevan

Publications and source records attributed to Michael A. Bevan.

4 recordsLinked to original sources

arXiv2023

Learning Effective SDEs from Brownian Dynamics Simulations of Colloidal Particles

We construct a reduced, data-driven, parameter dependent effective Stochastic Differential Equation (eSDE) for electric-field mediated colloidal crystallization using data obtained from Brownian Dynamics Simulations. We use Diffusion Maps (a manifold learning algorithm) to identify a set of useful latent observables. In this latent space we identify an eSDE using a deep learning architecture inspired by numerical stochastic integrators and compare it with the traditional Kramers-Moyal expansion estimation. We show that the obtained variables and the learned dynamics accurately encode the physics of the Brownian Dynamic Simulations. We further illustrate that our reduced model captures the dynamics of corresponding experimental data. Our dimension reduction/reduced model identification approach can be easily ported to a broad class of particle systems dynamics experiments/models.

math.DS↗

arXiv2021

Hierarchical deep reinforcement learning controlled three-dimensional navigation of microrobots in blood vessels

Designing intelligent microrobots that can autonomously navigate and perform instructed routines in blood vessels, a complex and crowded environment with obstacles including dense cells, different flow patterns and diverse vascular geometries, can offer enormous possibilities in biomedical applications. Here we report a hierarchical control scheme that enables a microrobot to efficiently navigate and execute customizable routines in blood vessels. The control scheme consists of two highly decoupled components: a high-level controller setting short-ranged dynamic targets to guide the microrobot to follow a preset path and a low-level deep reinforcement learning (DRL) controller responsible for maneuvering microrobots towards these dynamic guiding targets. The proposed DRL controller utilizes three-dimensional (3D) convolutional neural networks and is capable of learning control policy directly from a coarse raw 3D sensory input. In blood vessels with rich configurations of red blood cells and vessel geometry, the control scheme enables efficient navigation and faithful execution of instructed routines. The control scheme is also robust to adversarial perturbations including blood flows. This study provides a proof-of-principle for designing data-driven control systems for autonomous navigation in vascular networks; it illustrates the great potential of artificial intelligence for broad biomedical applications such as target drug delivery, blood clots clear, precision surgery, disease diagnosis, and more.

cond-mat.soft↗

arXiv2020

Micro/Nano Motor Navigation and Localization via Deep Reinforcement Learning

Efficient navigation and precise localization of Brownian micro/nano self-propelled motor particles within complex landscapes could enable future high-tech applications involving for example drug delivery, precision surgery, oil recovery, and environmental remediation. Here we employ a model-free deep reinforcement learning algorithm based on bio-inspired neural networks to enable different types of micro/nano motors to be continuously controlled to carry out complex navigation and localization tasks. Micro/nano motors with either tunable self-propelling speeds or orientations or both, are found to exhibit strikingly different dynamics. In particular, distinct control strategies are required to achieve effective navigation in free space and obstacle environments, as well as under time constraints. Our findings provide fundamental insights into active dynamics of Brownian particles controlled using artificial intelligence and could guide the design of motor and robot control systems with diverse application requirements.

cond-mat.soft↗

arXiv2019

Efficient Navigation of Colloidal Robots in an Unknown Environment via Deep Reinforcement Learning

Equipping active colloidal robots with intelligence such that they can efficiently navigate in unknown complex environments could dramatically impact their use in emerging applications like precision surgery and targeted drug delivery. Here we develop a model-free deep reinforcement learning that can train colloidal robots to learn effective navigation strategies in unknown environments with random obstacles. We show that trained robot agents learn to make navigation decisions regarding both obstacle avoidance and travel time minimization, based solely on local sensory inputs without prior knowledge of the global environment. Such agents with biologically inspired mechanisms can acquire competitive navigation capabilities in large-scale, complex environments containing obstacles of diverse shapes, sizes, and configurations. This study illustrates the potential of artificial intelligence in engineering active colloidal systems for future applications and constructing complex active systems with visual and learning capability.

cond-mat.soft↗