Search for: All records

This work presents a unique approach to the design, fabrication, and characterization of paper-based origami robotic systems consisting of stackable pneumatic actuators. These paper-based actuators (PBAs) use materials with high elastic modulus-to-mass ratios, accordion-like structures, and direct coupling with pneumatic pressure for extension and bending. The study contributes to the scientific and engineering understanding of foldable components under applied pneumatic pressure by constructing stretchable and flexible structures with intrinsically nonstretchable materials. Experiments showed that a PBA possesses a power-to-mass ratio greater than 80 W/kg, which is more than four times that of human muscle. This work also illustrates the stackability and functionality of PBAs by two prototypes: a parallel manipulator and a legged locomotor. The manipulator consisting of an array of PBAs can bend in a specific direction with the corresponding actuator inflated. In addition, the stacked actuators in the manipulator can rotate in opposite directions to compensate for relative rotation at the ends of each actuator to work in parallel and manipulate the platform. The locomotor rotates the PBAs to apply and release contact between the feet and the ground. Furthermore, a numerical model developed in this work predicts the mechanical performance of these inflatable actuators as a function of dimensional specifications and folding patterns. Overall, we use stacked origami actuators to implement functionalities of manipulation, gripping, and locomotion as conventional robotic systems. Future origami robots made of paper-like materials may be suitable for single use in contaminated or unstructured environments or low-cost educational materials. 

Aliasing refers to the phenomenon that high frequency signals degenerate into com- pletely different ones after sampling. It arises as a problem in the context of deep learning as downsampling layers are widely adopted in deep architectures to reduce parameters and computation. The standard solution is to apply a low-pass filter (e.g., Gaussian blur) before downsampling [37]. However, it can be suboptimal to apply the same filter across the entire content, as the frequency of feature maps can vary across both spatial locations and feature channels. To tackle this, we propose an adaptive content-aware low-pass filtering layer, which predicts separate filter weights for each spatial location and chan- nel group of the input feature maps. We investigate the effectiveness and generalization of the proposed method across multiple tasks including ImageNet classification, COCO instance segmentation, and Cityscapes semantic segmentation. Qualitative and quanti- tative results demonstrate that our approach effectively adapts to the different feature frequencies to avoid aliasing while preserving useful information for recognition. Code is available at https://maureenzou.github.io/ddac/.