The brain-inspired and event-driven Spiking Neural Network (SNN) aiming at mimicking the synaptic activity of biological neurons has received increasing attention. It transmits binary spike signals between network units when membrane potential exceeds firing threshold. This biomimetic mechanism SNN appears energy-efficiency with its power sparsity asynchronous operations on events. Unfortunately, propagation spikes, distribution will shift, leading to degeneration, saturation, gradient...

Learning to manipulate 3D objects in an interactive environment has been a challenging problem Reinforcement (RL). In particular, it is hard train policy that can generalize over with different semantic categories, diverse shape geometry and versatile functionality. this study, we focused on the contact information manipulation processes, proposed unified representation for critical interactions describe kinds of tasks. Specifically, take advantage generated during RL training process employ...

Achieving human-level dexterity is an important open problem in robotics. However, tasks of dexterous hand manipulation, even at the baby level, are challenging to solve through reinforcement learning (RL). The difficulty lies high degrees freedom and required cooperation among heterogeneous agents (e.g., joints fingers). In this study, we propose Bimanual Dexterous Hands Benchmark (Bi-DexHands), a simulator that involves two hands with tens bimanual manipulation thousands target objects....

Recently, the neuromorphic vision sensor has received more and interest. However, data consists of asynchronous event spikes, which makes it difficult to construct a big benchmark train power general neural network model, thus limiting understanding for "unseen" objects by deep learning. While frame image, since training can be obtained easily, zero-shot few-shot learning task via large Contrastive Vision-Language Pre-training (CLIP) is pre-trained large-scale image-text pairs in 2D, have...

The Spiking Neural Network (SNN), as one of the biologically inspired neural network infrastructures, has drawn increasing attention recently. It adopts binary spike activations to transmit information, thus multiplications and weights can be substituted by additions, which brings high energy efficiency. However, in paper, we theoretically experimentally prove that activation map cannot carry enough causing information loss resulting accuracy decreasing. To handle problem, propose a ternary...

Achieving accurate and generalized autonomous navigation in unknown environments poses a significant challenge robotics artificial intelligence. Animals exhibits superlative capabilities by combining the representation of internal neurals sensory cues self-motion external information. This paper proposes brain-inspired method based upon spiking neural networks (SNN) reinforcement learning, integrated with lidar system that serves as local environment explorer, which realizes high performance...

Obtaining the land vehicle's position, azimuth and attitudes autonomously accurately is significant for vehicle-based weapon's combat effectiveness. High performance inertial navigation system (INS) can provide high precision reference information, however, it usually very expensive. In addition, due to gyroscope drifts accelerometer biases, INS errors would diverge with time. To solve problem, this paper proposed an integrated positioning orientation method based on FOG single-axis...

Recently, Spiking Neural Networks (SNNs), enjoying extreme energy efficiency, have drawn much research attention on 2D visual recognition and shown gradually increasing application potential. However, it still remains underexplored whether SNNs can be generalized to 3D recognition. To this end, we present PointNet in the paper, first spiking neural model for efficient deep learning point clouds. We discover that two huge obstacles limiting of clouds are: intrinsic optimization obstacle...

The modern machine learning-based technologies have shown considerable potential in automatic radar scene understanding. Among these efforts, semantic segmentation (RSS) can provide more refined and detailed information including the moving objects background clutters within effective receptive field of radar. Motivated by success convolutional networks various visual computing tasks, also been introduced to solve RSS task. However, neither regular convolution operation nor modified ones are...

As one of the energy-efficient alternatives conventional neural networks (CNNs), spiking (SNNs) have gained more and interest recently. To train deep models, some effective batch normalization (BN) techniques are proposed in SNNs. All these BNs suggested to be used after convolution layer as usually doing CNNs. However, neuron is much complex with spatio-temporal dynamics. The regulated data flow BN will disturbed again by membrane potential updating operation before firing function, i.e.,...

Spiking neural networks (SNNs) have attracted intensive attention as a promising energy-efficient alternative to conventional artificial (ANNs) recently, which could transmit information in form of binary spikes rather than continuous activations thus the multiplication activation and weight be replaced by addition save energy. However, spike representation will sacrifice expression performance SNNs lead accuracy degradation compared with ANNs. Considering improving feature is beneficial...

Humans throw and catch objects all the time. However, such a seemingly common skill introduces lot of challenges for robots to achieve: The need operate dynamic actions at high-speed, collaborate precisely, interact with diverse objects. In this paper, we design system two multi-finger hands attached robot arms solve problem. We train our using Multi-Agent Reinforcement Learning in simulation perform Sim2Real transfer deploy on real robots. To overcome gap, provide multiple novel algorithm...

Rotation modulation technology could effectively improve the accuracy of inertial navigation system (INS) by compensating for biases sensors automatically. However, carrier angular motion and rotation control error reduce effect then decrease accuracy. To address this problem, single-axis INS, a novel scheme is presented. The employs fiber optic gyros to measurement unit (IMU) velocity so that INS with both azimuth insulation functions. Furthermore, in order error, study adopts two ways:...

Spiking Neural Networks (SNNs) as one of the biology-inspired models have received much attention recently. It can significantly reduce energy consumption since they quantize real-valued membrane potentials to 0/1 spikes transmit information thus multiplications activations and weights be replaced by additions when implemented on hardware. However, this quantization mechanism will inevitably introduce error, causing catastrophic loss. To address error problem, we propose a regularizing...

As one of the energy-efficient alternatives conventional neural networks (CNNs), spiking (SNNs) have gained more and interest recently. To train deep models, some effective batch normalization (BN) techniques are proposed in SNNs. All these BNs suggested to be used after convolution layer as usually doing CNNs. However, neuron is much complex with spatio-temporal dynamics. The regulated data flow BN will disturbed again by membrane potential updating operation before firing function, i.e.,...

The Spiking Neural Network (SNN) has attracted more and attention recently. It adopts binary spike signals to transmit information. Benefitting from the information passing paradigm of SNNs, multiplications activations weights can be replaced by additions, which are energy-efficient. However, its ``Hard Reset" mechanism for firing activity would ignore difference among membrane potentials when potential is above threshold, causing loss. Meanwhile, quantifying 0/1 spikes at instants will...

ADVERTISEMENT RETURN TO ISSUEPREVArticleNEXTHydrophobic interaction of ionenes in aqueous solutionZhuomei Li, Xuexin Zhang, Yuanpei Chen, and Yuanzhen ZhongCite this: Macromolecules 1992, 25, 1, 450–453Publication Date (Print):January 1992Publication History Published online1 May 2002Published inissue 1 January 1992https://pubs.acs.org/doi/10.1021/ma00027a070https://doi.org/10.1021/ma00027a070research-articleACS PublicationsRequest reuse permissionsArticle Views90Altmetric-Citations21LEARN...

Intersection-over-Union (IoU) is the most popular metric to evaluate regression performance in 3D object detection. Recently, there are also some methods applying IoU optimization of bounding box regression. However, we demonstrate through experiments and mathematical proof that loss suffers from abnormal gradient w.r.t. angular error scale, which further leads slow convergence suboptimal process, respectively. In this paper, propose a Gradient-Corrected (GCIoU) achieve fast accurate...

Spiking Neural Networks (SNNs) as one of the biology-inspired models have received much attention recently. It can significantly reduce energy consumption since they quantize real-valued membrane potentials to 0/1 spikes transmit information thus multiplications activations and weights be replaced by additions when implemented on hardware. However, this quantization mechanism will inevitably introduce error, causing catastrophic loss. To address error problem, we propose a regularizing...