Nature Machine Intelligence

Journal: Nature Machine Intelligence

Abbreviation

Nat Mach Intell

Publisher

Nature

ISSN

2522-5839

Show all metadata

Search Results

Publications 1 - 10 of 46

Autonomous 3D positional control of a magnetic microrobot using reinforcement learning
Abbasi, Sarmad Ahmad; Ahmed, Awais; Noh, Seungmin; et al. (2024)
Nature Machine Intelligence
Magnetic microrobots have shown promise in the field of biomedical engineering, facilitating precise drug delivery, non-invasive diagnosis and cell-based therapy. Current techniques for controlling the motion of such microrobots rely on the assumption of homogenous magnetic fields and are significantly influenced by a microrobot’s properties and surrounding environment. These strategies lack a sense of generality and adaptability when changing the environment or microrobot and exhibit a moderate delay due to independent control of the electromagnetic actuation system and microrobot’s position. To address these issues, we propose a machine learning-based positional control of magnetic microrobots via gradient fields generated by electromagnetic coils. We use reinforcement learning and a gradual training approach to control the three-dimensional position of a microrobot within a defined working area by directly managing the coil currents. We develop a simulation environment for initial exploration to reduce the overall training time. After simulation training, the learning process is transferred to a physical electromagnetic actuation system that reflects real-world intricacies. We compare our method to conventional proportional-integral-derivative control; our system is more accurate and efficient. The proposed method was combined with path planning algorithms to allow fully autonomous control. The presented approach is an alternative to complex mathematical models, which are sensitive to variations in microrobot design, the environment and the nonlinearity of magnetic systems.
Deep variational network for rapid 4D flow MRI reconstruction
Vishnevskiy, Valery; Walheim, Jonas; Kozerke, Sebastian (2020)
Nature Machine Intelligence
Phase-contrast magnetic resonance imaging (MRI) provides time-resolved quantification of blood flow dynamics that can aid clinical diagnosis. Long in vivo scan times due to repeated three-dimensional (3D) volume sampling over cardiac phases and breathing cycles necessitate accelerated imaging techniques that leverage data correlations. Standard compressed sensing reconstruction methods require tuning of hyperparameters and are computationally expensive, which diminishes the potential reduction of examination times. We propose an efficient model-based deep neural reconstruction network and evaluate its performance on clinical aortic flow data. The network is shown to reconstruct undersampled 4D flow MRI data in under a minute on standard consumer hardware. Remarkably, the relatively low amounts of tunable parameters allowed the network to be trained on images from 11 reference scans while generalizing well to retrospective and prospective undersampled data for various acceleration factors and anatomies.
Neuromorphic visual scene understanding with resonator networks
Renner, Alpha; Supic, Lazar; Danielescu, Andreea; et al. (2024)
Nature Machine Intelligence
Analysing a visual scene by inferring the configuration of a generative model is widely considered the most flexible and generalizable approach to scene understanding. Yet, one major problem is the computational challenge of the inference procedure, involving a combinatorial search across object identities and poses. Here we propose a neuromorphic solution exploiting three key concepts: (1) a computational framework based on vector symbolic architectures (VSAs) with complex-valued vectors, (2) the design of hierarchical resonator networks to factorize the non-commutative transforms translation and rotation in visual scenes and (3) the design of a multi-compartment spiking phasor neuron model for implementing complex-valued resonator networks on neuromorphic hardware. The VSA framework uses vector binding operations to form a generative image model in which binding acts as the equivariant operation for geometric transformations. A scene can therefore be described as a sum of vector products, which can then be efficiently factorized by a resonator network to infer objects and their poses. The hierarchical resonator network features a partitioned architecture in which vector binding is equivariant for horizontal and vertical translation within one partition and for rotation and scaling within the other partition. The spiking neuron model allows mapping the resonator network onto efficient and low-power neuromorphic hardware. Our approach is demonstrated on synthetic scenes composed of simple two-dimensional shapes undergoing rigid geometric transformations and colour changes. A companion paper demonstrates the same approach in real-world application scenarios for machine vision and robotics.
Model-based reinforcement learning for ultrasound-driven autonomous microrobots
Medany, Mahmoud; Piglia, Lorenzo; Achenbach, Liam; et al. (2025)
Nature Machine Intelligence
Reinforcement learning is emerging as a powerful tool for microrobots control, as it enables autonomous navigation in environments where classical control approaches fall short. However, applying reinforcement learning to microrobotics is difficult due to the need for large training datasets, the slow convergence in physical systems and poor generalizability across environments. These challenges are amplified in ultrasound-actuated microrobots, which require rapid, precise adjustments in high-dimensional action space, which are often too complex for human operators. Addressing these challenges requires sample-efficient algorithms that adapt from limited data while managing complex physical interactions. To meet these challenges, we implemented model-based reinforcement learning for autonomous control of an ultrasound-driven microrobot, which learns from recurrent imagined environments. Our non-invasive, AI-controlled microrobot offers precise propulsion and efficiently learns from images in data-scarce environments. On transitioning from a pretrained simulation environment, we achieved sample-efficient collision avoidance and channel navigation, reaching a 90% success rate in target navigation across various channels within an hour of fine-tuning. Moreover, our model initially generalized successfully in 50% of tasks in new environments, improving to over 90% with 30 min of further training. We further demonstrated real-time manipulation of microrobots in complex vasculatures under both static and flow conditions, thus underscoring the potential of AI to revolutionize microrobotics in biomedical applications.
Bioinspired acousto-magnetic microswarm robots with upstream motility
Ahmed, Daniel; Sukhov, Alexander; Hauri, David; et al. (2021)
Nature Machine Intelligence
The ability to propel against flows, that is, to perform positive rheotaxis, can provide exciting opportunities for applications in targeted therapeutics and non-invasive surgery. So far no biocompatible technologies exist for navigating microparticles upstream when they are in a background fluid flow. Inspired by many naturally occurring microswimmers—such as bacteria, spermatozoa and plankton—that utilize the no-slip boundary conditions of the wall to exhibit upstream propulsion, here we report on the design and characterization of self-assembled microswarms that can execute upstream motility in a combination of external acoustic and magnetic fields. Both acoustic and magnetic fields are safe to humans, non-invasive, can penetrate deeply into the human body and are well-developed in clinical settings. The combination of both fields can overcome the limitations encountered by single actuation methods. The design criteria of the acoustically induced reaction force of the microswarms, which is needed to perform rolling-type motion, are discussed. We show quantitative agreement between experimental data and our model that captures the rolling behaviour. The upstream capability provides a design strategy for delivering small drug molecules to hard-to-reach sites and represents a fundamental step towards the realization of micro- and nanosystem navigation against the blood flow.
The rise of robots in surgical environments during COVID-19
Zemmar, Ajmal; Lozano, Andres M.; Nelson, Bradley J. (2020)
Nature Machine Intelligence
The COVID-19 pandemic has changed our world and impacted multiple layers of our society. All frontline workers and in particular those in direct contact with patients have been exposed to major risk. To mitigate pathogen spread and protect healthcare workers and patients, medical services have been largely restricted, including cancellation of elective surgeries, which has posed a substantial burden for patients and immense economic loss for various hospitals. The integration of a robot as a shielding layer, physically separating the healthcare worker and patient, is a powerful tool to combat the omnipresent fear of pathogen contamination and maintain surgical volumes. In this Perspective, we outline detailed scenarios in the pre-, intra- and postoperative care, in which the use of robots and artificial intelligence can mitigate infectious contamination and aid patient management in the surgical environment during times of immense patient influx. We also discuss cost-effectiveness and benefits of surgical robotic systems beyond their use in pandemics. The current pandemic creates unprecedented demands for hospitals. Digitization and machine intelligence are gaining significance in healthcare to combat the virus. Their legacy may well outlast the pandemic and revolutionize surgical performance and management. © 2020 Springer Nature Limited.
Automated de novo molecular design by hybrid machine intelligence and rule-driven chemical synthesis
Button, Alexander; Merk, Daniel; Hiss, Jan A.; et al. (2019)
Nature Machine Intelligence
Chemical creativity in the design of new synthetic chemical entities (NCEs) with drug-like properties has been the domain of medicinal chemists. Here, we explore the capability of a chemistry-savvy machine intelligence to generate synthetically accessible molecules. DINGOS (design of innovative NCEs generated by optimization strategies) is a virtual assembly method that combines a rule-based approach with a machine learning model trained on successful synthetic routes described in chemical patent literature. This unique combination enables a balance between ligand-similarity-based generation of innovative compounds by scaffold hopping and the forward-synthetic feasibility of the designs. In a prospective proof-of-concept application, DINGOS successfully produced sets of de novo designs for four approved drugs that were in agreement with the desired structural and physicochemical properties. Target prediction indicated more than 50% of the designs to be biologically active. Four selected computer-generated compounds were successfully synthesized in accordance with the synthetic route proposed by DINGOS. The results of this study demonstrate the capability of machine learning models to capture implicit chemical knowledge from chemical reaction data and suggest feasible syntheses of new chemical matter.
Mind and machine in drug design
Schneider, Gisbert (2019)
Nature Machine Intelligence
Causal chambers as a real-world physical testbed for AI methodology
Gamella, Juan L.; Peters, Jonas; Bühlmann, Peter (2025)
Nature Machine Intelligence
In some fields of artificial intelligence, machine learning and statistics, the validation of new methods and algorithms is often hindered by the scarcity of suitable real-world datasets. Researchers must often turn to simulated data, which yields limited information about the applicability of the proposed methods to real problems. As a step forward, we have constructed two devices that allow us to quickly and inexpensively produce large datasets from non-trivial but well-understood physical systems. The devices, which we call causal chambers, are computer-controlled laboratories that allow us to manipulate and measure an array of variables from these physical systems, providing a rich testbed for algorithms from a variety of fields. We illustrate potential applications through a series of case studies in fields such as causal discovery, out-of-distribution generalization, change point detection, independent component analysis and symbolic regression. For applications to causal inference, the chambers allow us to carefully perform interventions. We also provide and empirically validate a causal model of each chamber, which can be used as ground truth for different tasks. The hardware and software are made open source, and the datasets are publicly available at causalchamber.org or through the Python package causalchamber.
Drug discovery with explainable artificial intelligence
Jiménez-Luna, José; Grisoni, Francesca; Schneider, Gisbert (2020)
Nature Machine Intelligence
Deep learning bears promise for drug discovery, including advanced image analysis, prediction of molecular structure and function, and automated generation of innovative chemical entities with bespoke properties. Despite the growing number of successful prospective applications, the underlying mathematical models often remain elusive to interpretation by the human mind. There is a demand for ‘explainable’ deep learning methods to address the need for a new narrative of the machine language of the molecular sciences. This Review summarizes the most prominent algorithmic concepts of explainable artificial intelligence, and forecasts future opportunities, potential applications as well as several remaining challenges. We also hope it encourages additional efforts towards the development and acceptance of explainable artificial intelligence techniques. © 2020, Springer Nature Limited

Publications 1 - 10 of 46

Journal: Nature Machine Intelligence

Abbreviation

Publisher

Journal Volumes

ISSN

Description

Filters

Search Results