Long Short Term Memory Model Based Position-Stiffness Control of Antagonistically Driven Twisted-Coiled Polymer Actuators Using Model Predictive Control

Luong, Tuan; Kim, Kihyeon; Seo, Sunae; Jeon, Jeongmin; Park, Chanyong; Doh, Myeongyun; Koo, Ja Choon; Choi, Hyouk Ryeol; Moon, Hyungpil

doi:10.1109/lra.2021.3068905

Cited by 20 publications

(16 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These would not be available during online prediction scenarios, as would be the case for feedback control. However, we consider these tests to be valid, since prior work has shown that temperature of SMA wires can be predicted reasonably well from Joule heating models [15] and that LSTM networks may also be able to predict temperature itself [16]. A possible direction for future work is therefore predicting temperature alongside bending angle.…”

Section: Resultsmentioning

confidence: 98%

“…in soft electrothermal actuator dynamics -to be carried through over a long period of time-dependence. Most importantly, LSTMs have been shown to accurately capture hysteresis in soft actuators [16] and other continuum robots [23] without the use of additional state variables.…”

Section: A Long Short-term Memory (Lstm) Networkmentioning

confidence: 99%

“…The fully-connected layer leads into a single node for the output actuation state: θ (t). This approach is a simpler sequential network than the heirarchical LSTMs of [16], which was used for similar electrothermal twistedcoil polymer actuators.…”

Section: B Network Structure and Implementationmentioning

confidence: 99%

“…1) is powered by antagonistic pair of SMA coils and contains SMA-mounted thermocouples, as well as a dielectric elastomer bend sensor embedded within a soft silicone rubber. As motivated by recent work in other electrothermal actuators [16], we employ a long short-term memory (LSTM) neural network to learn a mapping from sensor inputs, over time, to the total bending angle of the limb (Fig. 2(A)).…”

Section: Introductionmentioning

confidence: 99%

“…This article focuses on the modeling of our limb, in-situ, with no external sensing, analyzing the effects of different sensor data as well as different motions on learned predictions. In comparison, state-of-the-art hysteresis modeling in electrothermal actuators focuses primarily on a single type of motion with a fixed set of external sensors [9], [10], [16]. Since many untethered SMA-actuated soft robots do not include thermal sensing [14] due to design challenges [17], we train networks both with and without that data and quantify the prediction quality that is possible given different sensors.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

In-Situ Sensing and Dynamics Predictions for Electrothermally-Actuated Soft Robot Limbs

Sabelhaus,

Mehta,

Wertz

et al. 2021

Preprint

View full text Add to dashboard Cite

Modeling the dynamics of soft robot limbs with electrothermal actuators is generally challenging due to thermal and mechanical hysteresis and the complex physical interactions that can arise during robot operation. This article proposes a neural network based on long short-term memory (LSTM) to address these challenges in actuator modeling. A planar soft limb, actuated by a pair of shape memory alloy (SMA) coils and containing embedded sensors for temperature and angular deflection, is used as a test platform. Data from this robot are used to train LSTM neural networks, using different combinations of sensor data, to model both unidirectional (one SMA) and bidirectional (both SMAs) motion. Open-loop rollout results show that the learned model is able to predict motions over extraordinarily long open-loop timescales (10 minutes) with little drift. Prediction errors are on the order of the soft deflection sensor's accuracy, even when using only the actuator's pulse width modulation inputs for learning. These LSTM models can be used in-situ, without extensive sensing, helping to bring soft electrothermally-actuated robots into practical application.

show abstract

Section: Resultsmentioning

confidence: 98%

Section: A Long Short-term Memory (Lstm) Networkmentioning

confidence: 99%

Section: B Network Structure and Implementationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

In-Situ Sensing and Dynamics Predictions for Electrothermally-Actuated Soft Robot Limbs

Sabelhaus,

Mehta,

Wertz

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

Hysteresis Modeling of Twisted-Coiled Polymer Actuators Using Long Short Term Memory Networks

Luong

Seo

Kim

et al. 2021

Advances in Asian Mechanism and Machine Science

View full text Add to dashboard Cite

Accelerating the design and development of polymeric materials via deep learning: Current status and future challenges

Chen

et al. 2023

APL Machine Learning

View full text Add to dashboard Cite

The design and development of polymeric materials have been a hot domain for decades. However, traditional experiments and molecular simulations are time-consuming and labor-intensive, which no longer meet the requirements of new materials development. With the rapid advances of artificial intelligence and materials informatics, machine learning algorithms are increasingly applied in materials science, aiming to shorten the development period of new materials. With the evolution of polymeric materials, the structure of polymers has become more and more complex. Traditional machine learning algorithms often do not perform satisfactorily when dealing with complex data. Presently, deep learning algorithms, including deep neural networks, convolutional neural networks, generative adversarial networks, recurrent neural networks, and graph neural networks, show their uniquely excellent learning capabilities for large and complex data, which will be a powerful tool for the design and development of polymeric materials. This Review introduces principles of several currently popular deep learning algorithms and discusses their multiple applications in the materials field. Applications range from property prediction and molecular generation at the molecular level to structure identification and material synthesis in polymers. Finally, future challenges and opportunities for the application of deep learning in polymeric materials are discussed.

show abstract

Long Short Term Memory Model Based Position-Stiffness Control of Antagonistically Driven Twisted-Coiled Polymer Actuators Using Model Predictive Control

Cited by 20 publications

References 18 publications

In-Situ Sensing and Dynamics Predictions for Electrothermally-Actuated Soft Robot Limbs

In-Situ Sensing and Dynamics Predictions for Electrothermally-Actuated Soft Robot Limbs

Hysteresis Modeling of Twisted-Coiled Polymer Actuators Using Long Short Term Memory Networks

Accelerating the design and development of polymeric materials via deep learning: Current status and future challenges

Contact Info

Product

Resources

About