Biological fliers such as birds are known for their bounding flight maneuvers during which they fold their wings under their bodies to soar intermittently, or manipulate their inertial body dynamics to achieve challen...
Biological fliers such as birds are known for their bounding flight maneuvers during which they fold their wings under their bodies to soar intermittently, or manipulate their inertial body dynamics to achieve challenging trajectories. This combination of thrust vectoring and body control allows biological fliers to optimize for a wide number of objectives - ranging from aerodynamic drag minimization to maneuverability. However, combined posture control and thrust vectoring still remains largely unexplored in the aerial robotics community. In this paper, we use a dynamical model of an aerial robot with articulated thrusters to generate minimum time trajectories under spatially varying constraints. To this end, we formulate an optimal control problem that is solved numerically using trapezoidal collocation. Our results indicate that combining posture control and thrust vectoring can enable flying through narrow and spatially varying geometries as well as decreasing maneuver time by careful manipulation of shape inputs.
This study is about real-time 2D-3D human pose estimation without using the a priori structure of the skeleton and with a low number of parameters for regression tasks. Current graph convolution-based 3D human pose ta...
详细信息
For as long as light and matter have partnered, impurities have played a role in optical system performance. This remains generally true for photonic heat engines and especially the case for optical refrigeration. Bui...
详细信息
Principal Component Analysis (PCA) is a versatile Unsupervised Learning (UL) technique for reducing the dimensionality of datasets. As a result, PCA is widely used in consumer and research applications as a preprocess...
Principal Component Analysis (PCA) is a versatile Unsupervised Learning (UL) technique for reducing the dimensionality of datasets. As a result, PCA is widely used in consumer and research applications as a preprocessing tool for identifying important features prior to further analysis. In instances where on-site personnel or developers do not have the expertise to apply UL techniques, third party processors are frequently retained. However, the release of client or proprietary data poses a substantial security risk. This risk increases the regulatory and contractual burden on analysts when interacting with sensitive or classified information. Homomorphic Encryption (HE) cryptosystems are a novel family of encryption algorithms that permit approximate addition and multiplication on encrypted data. When applied to UL models, such as PCA, experts may apply their expertise while maintaining data privacy. In order to evaluate the potential application of Homomorphic Encryption, we implemented Principal Component Analysis using the Microsoft SEAL HE libraries. The resulting implementation was applied to the MNIST Handwritten dataset for feature reduction and image reconstruction. Based on our results, HE considerably increased the time required to process the dataset. However, the HE algorithm is still viable for non-real-time applications as it had an average pixel error of near-zero for all image reconstructions.
An introduction of intelligent interconnectivity for people and things has posed higher demands and more challenges for sixth-generation (6G) networks, such as high spectral efficiency and energy efficiency, ultra-low...
详细信息
Electric grid faults are increasingly the source of ignition for major wildfires. To reduce the likelihood of such ignitions in high risk situations, utilities use pre-emptive deenergization of power lines, commonly r...
详细信息
Decomposition and characterization of backscattered ultrasonic multiple interfering echoes is a critical step for analyzing the ultrasonic propagation path, propagation modes, coupling condition of the transducers and...
Decomposition and characterization of backscattered ultrasonic multiple interfering echoes is a critical step for analyzing the ultrasonic propagation path, propagation modes, coupling condition of the transducers and detecting defects within the propagation path. Chirplet Signal Decomposition (CSD) is an efficient way of analyzing ultrasonic echoes. However, CSD is computationally expensive and time-consuming for real-time signal processing applications. To address this problem, we present a System-on-Chip (SoC) implementation of the CSD algorithm, with the goal of speed optimization and real-time execution without sacrificing the accuracy of the signal decomposition and reconstruction. The implementation was tested on a Zynq Ultrascale+ series FPGA and achieved an echo estimation within two milliseconds.
This paper presents a system design for a smart bike helmet with multiple safety features that are intended to empower bicycle riders to proactively avoid potential sources of danger or injury. A Smart Sensor/Actuator...
This paper presents a system design for a smart bike helmet with multiple safety features that are intended to empower bicycle riders to proactively avoid potential sources of danger or injury. A Smart Sensor/Actuator Node (SSAN), driven by an Arduino Uno single-board microcontroller, contains input sensors and actuators to provide riders the ability to send and receive warnings promptly on their helmet. A vision Node, driven by an NVIDIA Jetson Nano and a cable pin-connected camera, executes AI object detection algorithms for any dangerous objects that are out of sight of the rider and sends alerts to the SSAN as needed. By combining safety features of the SSAN and vision Node while continuously sending data to an IoT-enabled backend web server, the safety operation of a typical bike ride can be substantially improved.
Unlearnable examples (UEs) seek to maximize testing error by making subtle modifications to training examples that are correctly labeled. Defenses against these poisoning attacks can be categorized based on whether sp...
Unlearnable examples (UEs) seek to maximize testing error by making subtle modifications to training examples that are correctly labeled. Defenses against these poisoning attacks can be categorized based on whether specific interventions are adopted during training. The first approach is training-time defense, such as adversarial training, which can mitigate poisoning effects but is computationally intensive. The other approach is pre-training purification, e.g., image short squeezing, which consists of several simple compressions but often encounters challenges in dealing with various UEs. Our work provides a novel disentanglement mechanism to build an efficient pre-training purification method. Firstly, we uncover rate-constrained variational autoencoders (VAEs), demonstrating a clear tendency to suppress the perturbations in UEs. We subsequently conduct a theoretical analysis for this phenomenon. Building upon these insights, we introduce a disentangle variational autoencoder (DVAE), capable of disentangling the perturbations with learnable class-wise embeddings. Based on this network, a two-stage purification approach is naturally developed. The first stage focuses on roughly eliminating perturbations, while the second stage produces refined, poison-free results, ensuring effectiveness and robustness across various scenarios. Extensive experiments demonstrate the remarkable performance of our method across CIFAR-10, CIFAR-100, and a 100-class ImageNet-subset. Code is available at https://***/yuyi-sd/D-VAE.
暂无评论