Learning Controllers for Continuum Soft Manipulators: Impact of Modeling and Looming Challenges DOI Creative Commons
Egidio Falotico, Enrico Donato, Carlo Alessi

et al.

Advanced Intelligent Systems, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 7, 2024

Soft manipulators, renowned for their compliance and adaptability, hold great promise in ability to engage safely effectively with intricate environments delicate objects. Nonetheless, controlling these soft systems presents distinctive hurdles owing nonlinear behavior complicated dynamics. Learning‐based controllers continuum manipulators offer a viable alternative model‐based approaches that may struggle account uncertainties variability materials, limiting effectiveness real‐world scenarios. can be trained through experience, exploiting various forward models differ physical assumptions, accuracy, computational cost. In this article, the key features of popular models, including geometrical, pseudo‐rigid, mechanical, or learned, are first summarized. Then, unique characterization learning‐based policies, emphasizing impact on control problem how state art evolves, is offered. This leads presented perspectives outlining current challenges future research trends machine‐learning applications within robotics.

Language: Английский

Data-Driven Methods Applied to Soft Robot Modeling and Control: A Review DOI Creative Commons
Zixi Chen, Federico Renda,

Alexia Le Gall

et al.

IEEE Transactions on Automation Science and Engineering, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 16

Published: Jan. 1, 2024

Soft robots show compliance and have infinite degrees of freedom. Thanks to these properties, such can be leveraged for surgery, rehabilitation, biomimetics, unstructured environment exploring, industrial grippers. In this case, they attract scholars from a variety areas. However, nonlinearity hysteresis effects also bring burden robot modeling. Moreover, following their flexibility adaptation, soft control is more challenging than rigid control. order model robots, large number data-driven methods are utilized in pairs or separately. This review first briefly introduces two foundations approaches, which physical models the Jacobian matrix, then summarizes three kinds statistical method, neural network, reinforcement learning. compares modeling controller features, e.g., dynamics, data requirement, target task, within among categories. Finally, we summarize features each method. A discussion about advantages limitations existing approaches presented, forecast future robots. website (https://sites.google.com/view/23zcb) built will updated frequently. Note Practitioners —This work motivated by need introducing parallel. Modeling play significant roles research, especially The nonlinear complex deformation necessitates specific approaches. We introduce state-of-the-art survey widely utilized. performance methods, considering some important like amount frequency, task. approach summarized, discuss possible area.

Language: Английский

Citations

6
An Overview of Data-Driven Paradigms for Identification and Control of Robotic Systems DOI

Chandan Kumar Sah,

Rajpal Singh, Jishnu Keshavan

et al.

Journal of the Indian Institute of Science, Journal Year: 2025, Volume and Issue: unknown

Published: March 12, 2025

Language: Английский

Citations

0
Programmable Shape‐Preserving Soft Robotics Arm via Multimodal Multistability DOI Creative Commons

Benyamin Shahryari,

Hossein Mofatteh,

Arian Sargazi

et al.

Advanced Functional Materials, Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 29, 2024

Abstract Inflatable multistable materials have significantly advanced the design of shape‐preserving soft robotic arms, offering substantial benefits in terms shape adaptability, energy efficiency, and safety, ensuring operational reliability even event sudden power loss. However, existing strategies for realizing arms often limit themselves to a single mode multistability, commonly with rotationally symmetric designs favoring extension stability asymmetric inducing bending stability. To address limitation, this study introduces pioneering platform termed multimodal multistability that utilizes geometrical frustration. A cylindrical cell, designed bistability, could achieve frustrated states by controlling cell multiple degrees freedom incorporated pneumatic actuator. This extends spectrum attainable stable trajectories while preserving essential attributes such as load‐bearability, programmability, reversibility changes. Leveraging system four pressure control, not only enables capturing previously unexplored configurations mechanical metastructures but also allows control their deformation modes. With applications spanning space exploration, medical instruments, rescue missions, promises unparalleled flexibility efficiency operation robots.

Language: Английский

Citations

3
A review on machine learning in flexible surgical and interventional robots: Where we are and where we are going DOI Creative Commons
Di Wu, Renchi Zhang, Ameya Pore

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 93, P. 106179 - 106179

Published: March 15, 2024

Minimally Invasive Procedures (MIPs) emerged as an alternative to more invasive surgical approaches, offering patient benefits such smaller incisions, less pain, and shorter hospital stay. In one class of MIPs, where natural body lumens or small incisions are used access deeper anatomical locations, Flexible Surgical Interventional Robots (FSIRs) catheters endoscopes widely used. Due their flexible compliant nature, FSIRs can be inserted via orifices then moved towards hard-to-reach targets perform interventional tasks. However, existing confronted with challenges in sensing, control, navigation. These issues stem from the robot's non-linear behavior intricate nature lumens, accurately modeling complex interactions disturbances proves exceptionally difficult. The rapid advances Machine Learning (ML) have facilitated widespread adoption ML techniques FSIRs. This article provides overview these efforts by first introducing a classification algorithms, including traditional methods modern Deep (DL) commonly Next, use algorithms is surveyed per sub-domain, namely for perception, modeling, Trends, popularity, strengths, and/or limitations different analyzed. roles that plays among tasks investigated described. Finally, discussions conducted on prospects MIPs.

Language: Английский

Citations

1
A Novel and Accurate BiLSTM Configuration Controller for Modular Soft Robots with Module Number Adaptability DOI
Zixi Chen, Matteo Bernabei, V. Mainardi

et al.

Soft Robotics, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 9, 2024

Modular soft robots (MSRs) exhibit greater potential for sophisticated tasks compared with single-module robots. However, the modular structure incurs complexity of accurate control and necessitates a strategy specifically In this article, we introduce data collection tailored MSR bidirectional long short-term memory (biLSTM) configuration controller capable adapting to varying module numbers. Simulation cable-driven real pneumatic have been included in experiments validate proposed approaches. Experimental results demonstrated that MSRs can explore larger space, thanks our method, be leveraged despite an increase or decrease number. By leveraging biLSTM, aim mimic physical MSRs, allowing adapt number change. Future work may include planning method bridges task, configuration, actuation spaces. We also integrate online components into controller.

Language: Английский

Citations

1
Learning Controllers for Continuum Soft Manipulators: Impact of Modeling and Looming Challenges DOI Creative Commons
Egidio Falotico, Enrico Donato, Carlo Alessi

et al.

Advanced Intelligent Systems, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 7, 2024

Soft manipulators, renowned for their compliance and adaptability, hold great promise in ability to engage safely effectively with intricate environments delicate objects. Nonetheless, controlling these soft systems presents distinctive hurdles owing nonlinear behavior complicated dynamics. Learning‐based controllers continuum manipulators offer a viable alternative model‐based approaches that may struggle account uncertainties variability materials, limiting effectiveness real‐world scenarios. can be trained through experience, exploiting various forward models differ physical assumptions, accuracy, computational cost. In this article, the key features of popular models, including geometrical, pseudo‐rigid, mechanical, or learned, are first summarized. Then, unique characterization learning‐based policies, emphasizing impact on control problem how state art evolves, is offered. This leads presented perspectives outlining current challenges future research trends machine‐learning applications within robotics.

Language: Английский

Citations

0