An official website of the United States government
Soft robot deformations are typically estimated using strain sensors to infer change from a nominal shape while taking a robot‐specific mechanical model into account. This approach performs poorly during buckling and when material properties change with time, and is untenable for shape‐changing robots that don't have a well‐defined resting (unactuated) shape. Herein, these limitations are overcome using stretchable shape sensing (S3) sheets that fuse orientation measurements to estimate 3D surface contours without making assumptions about the underlying robot geometry or material properties. The S3 sheets can estimate the shape of target objects to an accuracy of ≈3 mm for an 80 mm long sheet. The authors show the S3 sheets estimating their shape while being deformed in 3D space and also attached to the surface of a silicone three‐chamber pneumatic bladder, highlighting the potential for shape‐sensing sheets to be applied, removed, and reapplied to soft robots for shape estimation. Finally, the S3 sheets detecting their own stretch up to 30% strain is demonstrated. The approach introduced herein provides a generalized method for measuring the shape of objects without making strong assumptions about the objects, thus achieving a modular, mechanics model‐free approach to proprioception for wearable electronics and soft robotics.
more »
« less
Accounting for Hysteresis in the Forward Kinematics of Nonlinearly-Routed Tendon-Driven Continuum Robots via a Learned Deep Decoder Network
Tendon-driven continuum robots have been gaining popularity in medical applications due to their ability to curve around complex anatomical structures, potentially reducing the invasiveness of surgery. However, accurate modeling is required to plan and control the movements of these flexible robots. Physics-based models have limitations due to unmodeled effects, leading to mismatches between model prediction and actual robot shape. Recently proposed learning-based methods have been shown to overcome some of these limitations but do not account for hysteresis, a significant source of error for these robots. To overcome these challenges, we propose a novel deep decoder neural network that predicts the complete shape of tendon-driven robots using point clouds as the shape representation, conditioned on prior configurations to account for hysteresis. We evaluate our method on a physical tendon-driven robot and show that our network model accurately predicts the robot's shape, significantly outperforming a state-of-the-art physics-based model and a learning-based model that does not account for hysteresis.
more »
« less
- Award ID(s):
- 2133027
- PAR ID:
- 10547472
- Publisher / Repository:
- IEEE
- Date Published:
- Journal Name:
- IEEE Robotics and Automation Letters
- Volume:
- 9
- Issue:
- 11
- ISSN:
- 2377-3774
- Page Range / eLocation ID:
- 9263 to 9270
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Soft robots can undergo large elastic deformations and adapt to complex shapes. However, they lack the structural strength to withstand external loads due to the intrinsic compliance of fabrication materials (silicone or rubber). In this paper, we present a novel stiffness modulation approach that controls the robot’s stiffness on-demand without permanently affecting the intrinsic compliance of the elastomeric body. Inspired by concentric tube robots, this approach uses a Nitinol tube as the backbone, which can be slid in and out of the soft robot body to achieve robot pose or stiffness modulation. To validate the proposed idea, we fabricated a tendon-driven concentric tube (TDCT) soft robot and developed the model based on Cosserat rod theory. The model is validated in different scenarios by varying the joint-space tendon input and task-space external contact force. Experimental results indicate that the model is capable of estimating the shape of the TDCT soft robot with an average root-mean-square error (RMSE) of 0.90 (0.56% of total length) mm and average tip error of 1.49 (0.93% of total length) mm. Simulation studies demonstrate that the Nitinol backbone insertion can enhance the kinematic workspace and reduce the compliance of the TDCT soft robot by 57.7%. Two case studies (object manipulation and soft laparoscopic photodynamic therapy) are presented to demonstrate the potential application of the proposed design.more » « less
-
Advancements in robotics and AI have increased the demand for interactive robots in healthcare and assistive applications. However, ensuring safe and effective physical human-robot interactions (pHRIs) remains challenging due to the complexities of human motor communication and intent recognition. Traditional physics-based models struggle to capture the dynamic nature of human force interactions, limiting robotic adaptability. To address these limitations, neural networks (NNs) have been explored for force-movement intention prediction. While multi-layer perceptron (MLP) networks show potential, they struggle with temporal dependencies and generalization. Long Short-Term Memory (LSTM) networks effectively model sequential dependencies, while Convolutional Neural Networks (CNNs) enhance spatial feature extraction from human force data. Building on these strengths, this study introduces a hybrid LSTM-CNN framework to improve force-movement intention prediction, increasing accuracy from 69% to 86% through effective denoising and advanced architectures. The combined CNN-LSTM network proved particularly effective in handling individualized force-velocity relationships and presents a generalizable model paving the way for more adaptive strategies in robot guidance. These findings highlight the importance of integrating spatial and temporal modeling to enhance robot precision, responsiveness, and human-robot collaboration. Index Terms —- Physical Human-Robot Interaction, Intention Detection, Machine Learning, Long-Short Term Memory (LSTM)more » « less
-
Soft robots have recently drawn extensive attention thanks to their unique ability of adapting to complicated environments. Soft robots are designed in a variety of shapes of aiming for many different applications. However, accurate modelling and control of soft robots is still an open problem due to the complex robot structure and uncertain interaction with the environment. In fact, there is no unified framework for the modeling and control of generic soft robots. In this paper, we present a novel data-driven machine learning method for modeling a cable-driven soft robot. This machine learning algorithm, named deterministic learning (DL), uses soft robot motion data to train a radial basis function neural network (RBFNN). The soft robot motion dynamics are then guaranteed to be accurately identified, represented, and stored as an RBFNN model with converged constant neural network weights. To validate our method, We have built a simulated soft robot almost identical to our real inchworm soft robot, and we have tested the DL algorithm in simulation. Furthermore, a neural network weight combining technique is used which can extract and combine useful dynamics information from multiple robot motion trajectories.more » « less
-
Soft robots struggle with terrestrial locomotion due to their inherent lack of rigidity, specifically along axes not in line with the direction of actuation (side loads). We present a method for improved stiffness in a 3D printed, tendon-driven soft actuator. We show, both mathematically and experimentally, that our method leads to improved stiffness to these side loads. Additionally, we demonstrate the use of complex tendon routing schemes to achieve various trajectories with a single actuator morphology. Finally, we demonstrate that these two tendon routing strategies lead to improved locomotion speed and gait efficiency in a 3- legged, 3D printed, soft robot.more » « less
