ABSTRACT
This paper presents a benchmark analysis of NVIDIA Jetson platforms when operating deep learning-based 3D object detection frameworks. Three-dimensional (3D) object detection could be highly beneficial for the autonomous navigation of robotic platforms, such as autonomous vehicles, robots, and drones. Since the function provides one-shot inference that extracts 3D positions with depth information and the heading direction of neighboring objects, robots can generate a reliable path to navigate without collision. To enable the smooth functioning of 3D object detection, several approaches have been developed to build detectors using deep learning for fast and accurate inference. In this paper, we investigate 3D object detectors and analyze their performance on the NVIDIA Jetson series that contain an onboard graphical processing unit (GPU) for deep learning computation. Since robotic platforms often require real-time control to avoid dynamic obstacles, onboard processing with a built-in computer is an emerging trend. The Jetson series satisfies such requirements with a compact board size and suitable computational performance for autonomous navigation. However, a proper benchmark that analyzes the Jetson for a computationally expensive task, such as point cloud processing, has not yet been extensively studied. In order to examine the Jetson series for such expensive tasks, we tested the performance of all commercially available boards (i.e., Nano, TX2, NX, and AGX) with state-of-the-art 3D object detectors. We also evaluated the effect of the TensorRT library to optimize a deep learning model for faster inference and lower resource utilization on the Jetson platforms. We present benchmark results in terms of three metrics, including detection accuracy, frame per second (FPS), and resource usage with power consumption. From the experiments, we observe that all Jetson boards, on average, consume over 80% of GPU resources. Moreover, TensorRT could remarkably increase inference speed (i.e., four times faster) and reduce the central processing unit (CPU) and memory consumption in half. By analyzing such metrics in detail, we establish research foundations on edge device-based 3D object detection for the efficient operation of various robotic applications.
ABSTRACT
In this work, we introduce liquid metal patterned stretchable and soft capacitive sensor with enhanced dielectric properties enabled by graphite nanofiber (GNF) fillers dispersed in polydimethylsiloxane (PDMS) substrate. We oxidized gallium-based liquid metal that exhibited excellent wetting behavior on the surface of the composites to enable patterning of the electrodes by a facile stencil printing. The fluidic behavior of the liquid metal electrode and modulated dielectric properties of the composite (k = 6.41 ± 0.092@6 wt % at 1 kHz) was utilized to fabricate stretchable and soft capacitive sensor with ability to distinguish various hand motions.
ABSTRACT
Herein, ultrasoft and ultrastretchable wearable strain sensors enabled by liquid metal fillers in an elastic polymer are described. The wearable strain sensors that can change the effective resistance upon strains are prepared by mixing silicone elastomer with liquid metal (EGaIn, Eutectic gallium-indium alloy) fillers. While the silicone is mixed with the liquid metal by shear mixing, the liquid metal is rendered into small droplets stabilized by an oxide, resulting in a non-conductive liquid metal elastomer. To attain electrical conductivity, localized mechanical pressure is applied using a stylus onto the thermally cured elastomer, resulting in the formation of a handwritten conductive trace by rupturing the oxide layer of the liquid metal droplets and subsequent percolation. Although this approach has been introduced previously, the liquid metal dispersed elastomers developed here are compelling because of their ultra-stretchable (elongation at break of 4000%) and ultrasoft (Young's modulus of <0.1 MPa) mechanical properties. The handwritten conductive trace in the elastomers can maintain metallic conductivity when strained; however, remarkably, we observed that the electrical conductivity is anisotropic upon parallel and perpendicular strains to the conductive trace. This anisotropic conductivity of the liquid metal elastomer film can manipulate the locomotion of a robot by routing the power signals between the battery and the driving motor of a robot upon parallel and perpendicular strains to the hand-written circuit. In addition, the liquid metal dispersed elastomers have a high degree of deformation and adhesion; thus, they are suitable for use as a wearable sensor for monitoring various body motions.
ABSTRACT
In this work we describe a soft and ultrastretchable fiber with a magnetic liquid metal (MLM) core for electrical switches used in remote magnetic actuation. MLM was prepared by removing the oxide layer on the liquid metal and subsequent mixing with magnetic iron particles. We used SEBS (poly[styrene-b-(ethylene-co-butylene)-b-styrene]) and silicone to prepare stretchable elastic fibers. Once hollow elastic fibers form, MLM was injected into the core of the fiber at ambient pressure. The fibers are soft (Young's modulus of 1.6~4.4 MPa) and ultrastretchable (elongation at break of 600~5000%) while maintaining electrical conductivity and magnetic property due to the fluidic nature of the core. Magnetic strength of the fibers was characterized by measuring the maximum effective distance between the magnet and the fiber as a function of iron particle concentration in the MLM core and the polymeric shell. The MLM core facilitates the use of the fiber in electrical switches for remote magnetic actuation. This ultrastretchable and elastic fiber with MLM core can be used in soft robotics, and wearable and conformal electronics.
