Tactile hand gesture recognition is a crucial task for user control in the automotive sector, where Human-Machine Interactions (HMI) demand low latency and high energy efficiency. This study addresses the challenges of power-constrained edge training and inference by utilizing analog Resistive Random Access Memory (ReRAM) technology in conjunction with a real tactile hand gesture dataset. By optimizing the input space through a feature engineering strategy, we avoid relying on large-scale crossbar arrays, making the system more suitable for edge deployment. Through realistic hardware-aware simulations that account for device non-idealities derived from experimental data, we demonstrate the functionalities of our analog ReRAM-based analog in-memory computing for on-chip training, utilizing the state-of-the-art Tiki-Taka algorithm. Furthermore, we validate the classification accuracy of approximately 91.4% for post-deployment inference of hand gestures. The results highlight the potential of analog ReRAM technology and crossbar architecture with fully parallelized matrix computations for real-time HMI systems at the Edge.

Deep neural networks have recently advanced state-of-the-art in motion deblurring. However, non-uniform non-blind image deblurring has not been studied in depth. State-of-the-art methods shows improvement over conventional algorithms, but they are still not feasible for mobile deployment. Having informative prior information could improve performance of non-uniform deblurring. In this work, we propose a new deep framework that allows extracting spatially variant latent feature to Distortion Prior map from a pair of calibration sharp-blur images, without having to capture or model training dataset. We propose to use multi-scale Distortion Prior map that can fully utilize spatially variant information in the further restoration via multi-scale attention mechanism. Unlike prior art, we use image pyramid at decoder side, by fusing its fine level with coarse level of feature map via level attention and by injecting Distortion Prior at various resolution levels. Experiments show that proposed network outperforms state-of-the-art deblur networks both in terms of image quality and inference time. We demonstrate that proposed framework can successfully deblur non-uniform, non-blind applications, such as defocus blur removal. Being computationally efficient, it is feasible for mobile deployment.

The world largest mobile image sensor The large Tetra-Cell pixel guarantees DSR-like image with higher SNR in the dark and better resolution in the bright Dual Pixel Pro All directional phase detection auto focus function can improve the AF performance Smart ISO Pro High dynamic range solution without no motion artifact and merging function in AP Staggered HDR Provide cost-effective high dynamic range image without the motion blue The innovative low power scheme Provide the lower analog power consumption on 2.2V supply voltage Flexible high-speed interface MIPI combo can provide the user flexibility with EMI immunity

As the application of deep learning continues to grow, so does the amount of data used to make predictions. While traditionally big-data deep learning was constrained by computing performance and off-chip memory bandwidth, a new constraint has emerged: privacy. One solution is homomorphic encryption (HE). Applying HE to the client-cloud model allows cloud services to perform inferences directly on clients’ encrypted data. While HE can meet privacy constraints it introduces enormous computational challenges and remains impractically slow on current systems.This paper introduces Cheetah, a set of algorithmic and hardware optimizations for server-side HE DNN inference. Cheetah proposes HE-parameter tuning and operator scheduling optimizations, which together deliver up to $79 \times$ speedup over the state-of-the-art. However, HE inference still falls short of real-time inference speeds by nearly four orders of magnitude. Cheetah further proposes an accelerator architecture to understand the degree of speedup hardware can provide and whether it can bridge HE’s real-time performance gap. We evaluate several DNNs and find that privacy-preserving HE inference for ResNet50 can approach real-time speeds with a 587mm2accelerator dissipating 30W in 5nm.

The general approach to humanoid feet design considers the use of rectangular plate sole structures that are relatively stiff and compatible for flat terrain locomotion. Although this can be adequate the inability of these stiff feet to cope with small terrain irregularities, it makes them inappropriate for rough terrains. Toward improvement of humanoids feet this paper presents mechanism of new variable compliant humanoid feet which can provide functionality and adaptability to humanoids locomotion on uneven terrains. The proposed feet design introduces new toe mechanism in the feet with variable stiffness implemented using a leaf spring and rubber balls in series. We present the mechanism design and the implementation of the sensor, and discuss the estimation of variable stiffness range and coefficient of damping at sole. A prototype of the feet was built and experimental results are included to validate the feet design.

The deployment of robots to assist in environments hostile for humans during emergency scenarios require robots to demonstrate enhanced physical performance, that includes adequate power, adaptability and robustness to physical interactions and efficient operation. This work presents the design and development of the lower body of the new high performance humanoid WALK-MAN, a robot developed recently to assist in disaster response scenarios. The paper introduces the details of the WALK-MAN lower-body, highlighting the innovative design optimization features considered to maximize the leg performance. Starting from the general lower body specifications the objectives of the design and how they were addressed are introduced, including the selection of the leg kinematics, the arrangement of the actuators and their integration with the leg structure to maximize the range of motion, reduce the leg mass and inertia, and shape the leg mass distribution for better dynamic performance. Physical robustness is ensured with the integration of elastic transmission and impact energy absorbing covers. Experimental walking trials demonstrate the correct operation of the legs while executing a walking gait.

The work presents the development and evaluation of a novel foot sole for humanoid robots. For humanoid locomotion the foot sole is important for absorbing impacts. In contrast to the simple planar rubber pad foot sole that is conventionally used in humanoid robots this paper introduces a new foot sole design in which the dissipation of energy during collision is done effectively using a viscous air damping sole mechanism that provides better reduction of the ground impact forces. The paper describes the principle of the foot sole and provides details of its design and implementation. Experimental trials were performed with the child size humanoid robot, COMAN, wearing the proposed feet to validate their performance during landing and walking. The results demonstrate that the proposed new passive damping mechanism can reduce effectively the ground reaction impact forces and oscillations while maintaining the foot/body posture.