Abstract Modular active cell robots (MACROs) are a design paradigm for modular robotic hardware that uses only two components, namely actuators and passive compliant joints. Under the MACRO approach, a large number of actuators and joints are connected to create mesh-like cellular robotic structures that can be actuated to achieve large deformation and shape change. In this two-part paper, we study the importance of different possible mesh topologies within the MACRO framework. Regular and semi-regular tilings of the plane are used as the candidate mesh topologies and simulated using finite element analysis (FEA). In Part 1, we use FEA to evaluate their passive stiffness characteristics. Using a strain-energy method, the homogenized material properties (Young's modulus, shear modulus, and Poisson's ratio) of the different mesh topologies are computed and compared. The results show that the stiffnesses increase with increasing nodal connectivity and that stretching-dominated topologies have higher stiffness compared to bending-dominated ones. We also investigate the role of relative actuator-node stiffness on the overall mesh characteristics. This analysis shows that the stiffness of stretching-dominated topologies scales directly with their cross-section area whereas bending-dominated ones do not have such a direct relationship.
more »
« less
Performance Comparison of Miniaturized Isolation Transformer Topologies
Seven different miniaturized low-power isolation transformer topologies are analyzed using FEA simulation. Performance optimizations of fixed-area transformers with different operation frequencies and heights, with and without magnetic cores, are performed. The comparison of optimization results reveals performance potential and trade-offs of the different topologies under different restrictions. More than 1 W of power transfer is possible at more than 85% efficiency in 1 mm2 of footprint area for several different topologies.
more »
« less
- Award ID(s):
- 1822140
- PAR ID:
- 10437948
- Date Published:
- Journal Name:
- 2023 IEEE Applied Power Electronics Conference and Exposition (APEC)
- Page Range / eLocation ID:
- 2654 to 2660
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
In this paper, we investigate the design of pennate topology fluidic artificial muscle bundles under spatial constraints. Soft fluidic actuators are of great interest to roboticists and engineers, due to their potential for inherent compliance and safe human–robot interaction. McKibben fluidic artificial muscles are an especially attractive type of soft fluidic actuator, due to their high force-to-weight ratio, inherent flexibility, inexpensive construction, and muscle-like force-contraction behavior. The examination of natural muscles has shown that those with pennate fiber topology can achieve higher output force per geometric cross-sectional area. Yet, this is not universally true for fluidic artificial muscle bundles, because the contraction and rotation behavior of individual actuator units (fibers) are both key factors contributing to situations where bipennate muscle topologies are advantageous, as compared to parallel muscle topologies. This paper analytically explores the implications of pennation angle on pennate fluidic artificial muscle bundle performance with spatial bounds. A method for muscle bundle parameterization as a function of desired bundle spatial envelope dimensions has been developed. An analysis of actuation performance metrics for bipennate and parallel topologies shows that bipennate artificial muscle bundles can be designed to amplify the muscle contraction, output force, stiffness, or work output capacity, as compared to a parallel bundle with the same envelope dimensions. In addition to quantifying the performance trade space associated with different pennate topologies, analyzing bundles with different fiber boundary conditions reveals how bipennate fluidic artificial muscle bundles can be designed for extensile motion and negative stiffness behaviors. This study, therefore, enables tailoring the muscle bundle parameters for custom compliant actuation applications.more » « less
-
As the machine learning and systems communities strive to achieve higher energy-efficiency through custom deep neural network (DNN) accelerators, varied precision or quantization levels, and model compression techniques, there is a need for design space exploration frameworks that incorporate quantization-aware processing elements into the accelerator design space while having accurate and fast power, performance, and area models. In this work, we present QUIDAM , a highly parameterized quantization-aware DNN accelerator and model co-exploration framework. Our framework can facilitate future research on design space exploration of DNN accelerators for various design choices such as bit precision, processing element type, scratchpad sizes of processing elements, global buffer size, number of total processing elements, and DNN configurations. Our results show that different bit precisions and processing element types lead to significant differences in terms of performance per area and energy. Specifically, our framework identifies a wide range of design points where performance per area and energy varies more than 5 × and 35 ×, respectively. With the proposed framework, we show that lightweight processing elements achieve on par accuracy results and up to 5.7 × more performance per area and energy improvement when compared to the best INT16 based implementation. Finally, due to the efficiency of the pre-characterized power, performance, and area models, QUIDAM can speed up the design exploration process by 3-4 orders of magnitude as it removes the need for expensive synthesis and characterization of each design.more » « less
-
This paper presents two novel single-phase resonant multilevel modular boost inverters based on resonant switched capacitor cells and a partial power processed voltage regulator. Compared with other multilevel boost inverters applied in PV systems, one remarkable advantage of the proposed topologies is that the bulky AC filtering inductor is replaced by a smaller-size one in the partial power processed buck converter. Constant duty cycle PWM method is attractive for the multilevel inverter controller design. GaN Enhancement Mode Power Transistors help both the modular resonant switched capacitor cells and the full-bridge unfolder be realized in a small size with high power density. The clamp capacitors in the resonant switched capacitor cells effectively alleviate the switch voltage spikes. These two inverter topologies are analyzed and simulated in PLECS. Simulation results verify the validity of boost inverter function. Stress analysis shows that the inverter has relatively small total normalized switch conduction power stress and total normalized switch stress ratio. Relative total semiconductor chip area comparison results reflect that the proposed topology achieves more efficient semiconductor utilization compared with typical non-resonant multilevel modular switched capacitor boost inverters. Test results indicate that the proposed topology can be used for single-phase non-isolated PV boost inverter applications with small ground leakage current, high voltage conversion ratio, small volume and potential high efficiency.more » « less
-
Noisy labels are inevitable in large real-world datasets. In this work, we explore an area understudied by previous works --- how the network's architecture impacts its robustness to noisy labels. We provide a formal framework connecting the robustness of a network to the alignments between its architecture and target/noise functions. Our framework measures a network's robustness via the predictive power in its representations --- the test performance of a linear model trained on the learned representations using a small set of clean labels. We hypothesize that a network is more robust to noisy labels if its architecture is more aligned with the target function than the noise. To support our hypothesis, we provide both theoretical and empirical evidence across various neural network architectures and different domains. We also find that when the network is well-aligned with the target function, its predictive power in representations could improve upon state-of-the-art (SOTA) noisy-label-training methods in terms of test accuracy and even outperform sophisticated methods that use clean labels.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/APEC43580.2023.10131134
