Dexterous grasping remains a fundamental yet challenging problem in robotics. A general-purpose robot must be capable of grasping diverse objects in arbitrary scenarios. However, existing research typically relies on restrictive assumptions, such as single-object settings or limited environments, showing constrained generalization. We present DexGraspVLA, a hierarchical framework for robust generalization in language-guided general dexterous grasping and beyond. It utilizes a pre-trained Vision-Language model as the high-level planner and learns a diffusion-based low-level Action controller. The key insight to achieve generalization lies in iteratively transforming diverse language and visual inputs into domain-invariant representations via foundation models, where imitation learning can be effectively applied due to the alleviation of domain shift. Notably, our method achieves a 90+% dexterous grasping success rate under thousands of challenging unseen cluttered scenes. Empirical analysis confirms the consistency of internal model behavior across environmental variations, validating our design. DexGraspVLA also, for the first time, simultaneously demonstrates free-form long-horizon prompt execution, robustness to adversarial objects and human disturbance, and failure recovery. Extended application to nonprehensile grasping further proves its generality.

Yuanfei Wang, Xinju Huang, Fangwei Zhong, Yaodong Yang, Yizhou Wang, Yuanpei Chen, Hao Dong. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Generating realistic future urban remote sensing imagery is critical for visualizing potential urban changes and supporting related technical analysis within urban planning. Traditional 2D-assisted methods are inherently limited in synthesize vertical development and infrastructure evolution, as they rely on binary planning maps. To address these limitations, we propose a novel 2.5D-assisted future urban remote sensing image synthesis method, aimed at generating future urban layouts based on existing urban structures and 2.5D planning maps. Specifically, the 2.5D map is divided into construction and demolition components, which are then integrated with the existing layout images and the embedding of the corresponding text as conditions for our generative model. We further design two trainable cascaded gated attention layers that process these two conditions separately and embed them into the latent diffusion model (LDM). This approach allows our model to dynamically comprehend the planning design requirements for key areas, making adjustments to accommodate diverse demands. Compared to existing state-of-the-art (SoTA) methods, our approach effectively targets design requirements, enabling flexible modifications that involve new constructions and demolitions in relevant urban areas. Experimental results on the 3DCD dataset demonstrate that the images generated by our method retain high fidelity and exhibit strong consistency with the 2.5D planning map.

Transformers have demonstrated outstanding performance across a wide range of tasks, owing to their self-attention mechanism, but they are highly energy-consuming. Spiking Neural Networks have emerged as a promising energy-efficient alternative to traditional Artificial Neural Networks, leveraging event-driven computation and binary spikes for information transfer. The combination of Transformers’ capabilities with the energy efficiency of SNNs offers a compelling opportunity. This paper addresses the challenge of adapting the self-attention mechanism of Transformers to the spiking paradigm by introducing a novel approach: Accurate Addition-Only Spiking Self-Attention (A2OS2A). Unlike existing methods that rely solely on binary spiking neurons for all components of the self-attention mechanism, our approach integrates binary, ReLU, and ternary spiking neurons. This hybrid strategy significantly improves accuracy while preserving non-multiplicative computations. Moreover, our method eliminates the need for softmax and scaling operations. Extensive experiments show that the A2OS2A-based Spiking Transformer outperforms existing SNN-based Transformers on several datasets, even achieving an accuracy of 78.66% on ImageNet-1K. Our work represents a significant advancement in SNN-based Transformer models, offering a more accurate and efficient solution for real-world applications.

Diffusion policies are widely adopted in complex visuomotor tasks for their ability to capture multimodal action distributions. However, the multiple sampling steps required for action generation significantly harm real-time inference efficiency, which limits their applicability in real-time decision-making scenarios. Existing acceleration techniques either require retraining or degrade performance under low sampling steps. Here we propose Falcon, which mitigates this speed-performance trade-off and achieves further acceleration. The core insight is that visuomotor tasks exhibit sequential dependencies between actions. Falcon leverages this by reusing partially denoised actions from historical information rather than sampling from Gaussian noise at each step. By integrating current observations, Falcon reduces sampling steps while preserving performance. Importantly, Falcon is a training-free algorithm that can be applied as a plug-in to further improve decision efficiency on top of existing acceleration techniques. We validated Falcon in 48 simulated environments and 2 real-world robot experiments. demonstrating a 2-7x speedup with negligible performance degradation, offering a promising direction for efficient visuomotor policy design.

The Spiking Neural Network (SNN), a biologically inspired neural network infrastructure, has garnered significant attention recently. SNNs utilize binary spike activations for efficient information transmission, replacing multiplications with additions, thereby enhancing energy efficiency. However, binary spike activation maps often fail to capture sufficient data information, resulting in reduced accuracy. To address this challenge, we advocate reversing the bit of the weight and activation, called \textbf{ReverB}, inspired by recent findings that highlight greater accuracy degradation from quantizing activations compared to weights. Specifically, our method employs real-valued spike activations alongside binary weights in SNNs. This preserves the event-driven and multiplication-free advantages of standard SNNs while enhancing the information capacity of activations. Additionally, we introduce a trainable factor within binary weights to adaptively learn suitable weight amplitudes during training, thereby increasing network capacity. To maintain efficiency akin to vanilla \textbf{ReverB}, our trainable binary weight SNNs are converted back to standard form using a re-parameterization technique during inference. Extensive experiments across various network architectures and datasets, both static and dynamic, demonstrate that our approach consistently outperforms state-of-the-art methods.

Autonomous learning of dexterous, long-horizon robotic skills has been a longstanding pursuit of embodied AI. Recent advances in robotic reinforcement learning (RL) have demonstrated remarkable performance and robustness in real-world visuomotor control tasks. However, applying RL in the real world faces challenges such as low sample efficiency, slow exploration, and significant reliance on human intervention. In contrast, simulators offer a safe and efficient environment for extensive exploration and data collection, while the visual sim-to-real gap, often a limiting factor, can be mitigated using real-to-sim techniques. Building on these, we propose SimLauncher, a novel framework that combines the strengths of real-world RL and real-to-sim-to-real approaches to overcome these challenges. Specifically, we first pre-train a visuomotor policy in the digital twin simulation environment, which then benefits real-world RL in two ways: (1) bootstrapping target values using extensive simulated demonstrations and real-world demonstrations derived from pre-trained policy rollouts, and (2) Incorporating action proposals from the pre-trained policy for better exploration. We conduct comprehensive experiments across multi-stage, contact-rich, and dexterous hand manipulation tasks. Compared to prior real-world RL approaches, SimLauncher significantly improves sample efficiency and achieves near-perfect success rates. We hope this work serves as a proof of concept and inspires further research on leveraging large-scale simulation pre-training to benefit real-world robotic RL.

Vision-language-action models (VLAs) show potential as generalist robot policies. However, these models pose extreme safety challenges during real-world deployment, including the risk of harm to the environment, the robot itself, and humans. *How can safety constraints be explicitly integrated into VLAs?* We address this by exploring an integrated safety approach (ISA), systematically **modeling** safety requirements, then actively **eliciting** diverse unsafe behaviors, effectively **constraining** VLA policies via safe reinforcement learning, and rigorously **assuring** their safety through targeted evaluations. Leveraging the constrained Markov decision process (CMDP) paradigm, ISA optimizes VLAs from a min-max perspective against elicited safety risks. Thus, policies aligned through this comprehensive approach achieve the following key features: (I) effective **safety-performance trade-offs**, reducing the cumulative cost of safety violations by 83.58\% compared to the state-of-the-art method, while also maintaining task success rate (+3.85\%). (II) strong **safety assurance**, with the ability to mitigate long-tail risks and handle extreme failure scenarios. (III) robust **generalization** of learned safety behaviors to various out-of-distribution perturbations. The effectiveness is evaluated on long-horizon mobile manipulation tasks.

Dexterous manipulation is critical for advancing robot capabilities in real-world applications, yet diverse and high-quality datasets remain scarce. Existing data collection methods either rely on human teleoperation or require significant human engineering, or generate data with limited diversity, which restricts their scalability and generalization. In this paper, we introduce DexFlyWheel, a scalable data generation framework that employs a self-improving cycle to continuously enrich data diversity. Starting from efficient seed demonstrations warmup, DexFlyWheel expands the dataset through iterative cycles. Each cycle follows a closed-loop pipeline that integrates Imitation Learning (IL), residual Reinforcement Learning (RL), rollout trajectory collection, and data augmentation. Specifically, IL extracts human-like behaviors from demonstrations, and residual RL enhances policy generalization. The learned policy is then used to generate trajectories in simulation, which are further augmented across diverse environments and spatial configurations before being fed back into the next cycle. Over successive iterations, a self-improving data flywheel effect emerges, producing datasets that cover diverse scenarios and thereby scaling policy performance. Experimental results demonstrate that DexFlyWheel generates over 2,000 diverse demonstrations across four challenging tasks. Policies trained on our dataset achieve an average success rate of 81.9\% on the challenge test sets and successfully transfer to the real world through digital twin, achieving a 78.3\% success rate on dual-arm lift tasks.

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 the multiplications of activations and weights can be substituted by additions, which brings high energy efficiency. However, in the paper, we theoretically and experimentally prove that the binary spike activation map cannot carry enough information, thus causing information loss and resulting in accuracy decreasing. To handle the problem, we propose a ternary spike neuron to transmit information. The ternary spike neuron can also enjoy the event-driven and multiplication-free operation advantages of the binary spike neuron but will boost the information capacity. Furthermore, we also embed a trainable factor in the ternary spike neuron to learn the suitable spike amplitude, thus our SNN will adopt different spike amplitudes along layers, which can better suit the phenomenon that the membrane potential distributions are different along layers. To retain the efficiency of the vanilla ternary spike, the trainable ternary spike SNN will be converted to a standard one again via a re-parameterization technique in the inference. Extensive experiments with several popular network structures over static and dynamic datasets show that the ternary spike can consistently outperform state-of-the-art methods. Our code is open-sourced at https://github.com/yfguo91/Ternary-Spike.

Spiking neural networks (SNNs) have attracted intensive attention as a promising energy-efficient alternative to conventional artificial neural networks (ANNs) recently, which could transmit information in form of binary spikes rather than continuous activations thus the multiplication of activation and weight could be replaced by addition to save energy. However, the binary spike representation form will sacrifice the expression performance of SNNs and lead to accuracy degradation compared with ANNs. Considering improving feature representation is beneficial to training an accurate SNN model, this paper focuses on enhancing the feature representation of the SNN. To this end, we establish a similarity-sensitive contrastive learning framework, where SNN could capture significantly more information from its ANN counterpart to improve representation by Mutual Information (MI) maximization with layer-wise sensitivity to similarity. In specific, it enriches the SNN’s feature representation by pulling the positive pairs of SNN's and ANN's feature representation of each layer from the same input samples closer together while pushing the negative pairs from different samples further apart. Experimental results show that our method consistently outperforms the current state-of-the-art algorithms on both popular non-spiking static and neuromorphic datasets.

Spiking neural networks (SNNs) have gained more and more interest as one of the energy-efficient alternatives of conventional artificial neural networks (ANNs). They exchange 0/1 spikes for processing information, thus most of the multiplications in networks can be replaced by additions. However, binary spike feature maps will limit the expressiveness of the SNN and result in unsatisfactory performance compared with ANNs. It is shown that a rich output feature representation, i.e., the feature vector before classifier) is beneficial to training an accurate model in ANNs for classification. We wonder if it also does for SNNs and how to improve the feature representation of the SNN. To this end, we materialize this idea in two special designed methods for SNNs. First, inspired by some ANN-SNN methods that directly copy-paste the weight parameters from trained ANN with light modification to homogeneous SNN can obtain a well-performed SNN, we use rich information of the weight parameters from the trained ANN counterpart to guide the feature representation learning of the SNN. In particular, we present the SNN's and ANN's feature representation from the same input to ANN's classifier to product SNN's and ANN's outputs respectively and then align the feature with the KL-divergence loss as in knowledge distillation methods, called L_ AF loss. It can be seen as a novel and effective knowledge distillation method specially designed for the SNN that comes from both the knowledge distillation and ANN-SNN methods. Various ablation study shows that the L_AF loss is more powerful than the vanilla knowledge distillation method. Second, we replace the last Leaky Integrate-and-Fire (LIF) activation layer as the ReLU activation layer to generate the output feature, thus a more powerful SNN with full-precision feature representation can be achieved but with only a little extra computation. Experimental results show that our method consistently outperforms the current state-of-the-art algorithms on both popular non-spiking static and neuromorphic datasets. We provide an extremely simple but effective way to train high-accuracy spiking neural networks.

The Spiking Neural Network (SNN) is a biologically inspired neural network infrastructure that has recently garnered significant attention. It utilizes binary spike activations to transmit information, thereby replacing multiplications with additions and resulting in high energy efficiency. However, training an SNN directly poses a challenge due to the undefined gradient of the firing spike process. Although prior works have employed various surrogate gradient training methods that use an alternative function to replace the firing process during back-propagation, these approaches ignore an intrinsic problem: gradient vanishing. To address this issue, we propose a shortcut back-propagation method in the paper, which advocates for transmitting the gradient directly from the loss to the shallow layers. This enables us to present the gradient to the shallow layers directly, thereby significantly mitigating the gradient vanishing problem. Additionally, this method does not introduce any burden during the inference phase. To strike a balance between final accuracy and ease of training, we also propose an evolutionary training framework and implement it by inducing a balance coefficient that dynamically changes with the training epoch, which further improves the network's performance. Extensive experiments conducted over static and dynamic datasets using several popular network structures reveal that our method consistently outperforms state-of-the-art methods.

Manipulating garments and fabrics has long been a critical endeavor in the development of home-assistant robots. However, due to complex dynamics and topological structures, garment manipulations pose significant challenges. Recent successes in reinforcement learning and vision-based methods offer promising avenues for learning garment manipulation. Nevertheless, these approaches are severely constrained by current benchmarks, which exhibit offer limited diversity of tasks and unrealistic simulation behavior. Therefore, we present GarmentLab, a content-rich benchmark and realistic simulation designed for deformable object and garment manipulation. Our benchmark encompasses a diverse range of garment types, robotic systems and manipulators. The abundant tasks in the benchmark further explores of the interactions between garments, deformable objects, rigid bodies, fluids, and human body. Moreover, by incorporating multiple simulation methods such as FEM and PBD, along with our proposed sim-to-real algorithms and real-world benchmark, we aim to significantly narrow the sim-to-real gap. We evaluate state-of-the-art vision methods, reinforcement learning, and imitation learning approaches on these tasks, highlighting the challenges faced by current algorithms, notably their limited generalization capabilities. Our proposed open-source environments and comprehensive analysis show promising boost to future research in garment manipulation by unlocking the full potential of these methods. We guarantee that we will open-source our code as soon as possible. You can watch the videos in supplementary files to learn more about the details of our work.

Radar signal interpretation plays a crucial role in remote detection and ranging. With the gradual display of the advantages of neural network technology in signal processing, learning-based radar signal interpretation is becoming a research hot-spot and made great progress. And since radar semantic segmentation (RSS) can provide more fine-grained target information, it has become a more concerned direction in this field. However, the temporal information, which is an important clue for analyzing radar data, has not been exploited sufficiently in present RSS frameworks. In this work, we propose a novel temporal information learning paradigm, i.e., data-driven temporal information aggregation with learned target-history relations. Following this idea, a flexible learning module, called Temporal Relation-Aware Module (TRAM) is carefully designed. TRAM contains two main blocks: i) an encoder for capturing the target-history temporal relations (TH-TRE) and ii) a learnable temporal relation attentive pooling (TRAP) for aggregating temporal information. Based on TRAM, an end-to-end Temporal-Aware RSS Network (TARSS-Net) is presented, which has outstanding performance on publicly available and our collected real-measured datasets. Code and supplementary materials are available at https://github.com/zlw9161/TARSS-Net.

Achieving human-level dexterity in robotics remains a critical open problem. Even simple dexterous manipulation tasks pose significant difficulties due to the high number of degrees of freedom and the need for cooperation among heterogeneous agents (e.g., finger joints). While some researchers have utilized reinforcement learning (RL) to control a single hand in manipulating objects, tasks that require coordinated bimanual cooperation are still under-explored due to the fewer suitable environments, which can result in difficulties and sub-optimal performance. To address these challenges, we introduce Bi-DexHands, a simulator with two dexterous hands featuring 20 bimanual manipulation tasks and thousands of target objects, designed to match various levels of human motor skills based on cognitive science research. We developed Bi-DexHands in Issac Gym, enabling highly efficient RL training at over 30,000 frames per second using a single NVIDIA RTX 3090. Based on Bi-DexHands, we present a comprehensive evaluation of popular RL algorithms in different settings, including single-agent/multi-agent RL, offline RL, multi-task RL, and meta RL. Our findings show that on-policy algorithms, such as PPO, can master simple manipulation tasks that correspond to those of 48-month-old babies, such as catching a flying object or opening a bottle. Furthermore, multi-agent RL can improve the ability to perform manipulations that require skilled bimanual cooperation, such as lifting a pot or stacking blocks. Despite achieving success in individual tasks, current RL algorithms struggle to learn multiple manipulation skills in most multi-task and few-shot learning scenarios. This highlights the need for further research and development within the RL community.

Recently, the neuromorphic vision sensor has received more and more interest. However, the neuromorphic data consists of asynchronous event spikes, which makes it difficult to construct a big benchmark to train a power general neural network model, thus limiting the neuromorphic data understanding for “unseen” objects by deep learning. While for the frame image, since the training data can be obtained easily, the zero-shot and few-shot learning for “unseen” task via the large Contrastive Vision-Language Pre-training (CLIP) model, which is pre-trained by large-scale image-text pairs in 2D, have shown inspirational performance. We wonder whether the CLIP could be transferred to neuromorphic data recognition to handle the “unseen” problem. To this end, we materialize this idea with NeuroCLIP in the paper. The NeuroCLIP consists of 2D CLIP and two specially designed modules for neuromorphic data understanding. First, an event-frame module that could convert the event spikes to the sequential frame image with a simple discrimination strategy. Second, an inter-timestep adapter, which is a simple fine-tuned adapter based on a spiking neural network (SNN) for the sequential features coming from the visual encoder of CLIP to improve the few-shot performance. Various experiments on neuromorphic datasets including N-MNIST, CIFAR10-DVS, and ES-ImageNet demonstrate the effectiveness of NeuroCLIP.

Reconstructing clear scene images in the presence of foreground occlusions remains a formidable challenge for cameras constrained by a single viewpoint. Synthetic aperture imaging (SAI) has emerged as a solution by integrating visual information from multiple viewpoints, thus overcoming occlusions. Event cameras, characterized by their high temporal resolution, exceptional dynamic range, low power consumption, and resistance to motion blur, offer a promising avenue for capturing intricate details of background objects within a limited range of motion. Leveraging these capabilities, several event-camera-based SAI methodologies have been proposed to effectively tackle dense occlusions. Despite these advancements, existing methodologies encounter obstacles stemming from the hybrid model they employ, comprising a spiking neural network (SNN) encoder and a convolutional neural network (CNN) decoder. Challenges include information degradation within the SNN encoder due to the quantization of full-precision data into binary spikes, as well as insufficient training data for the CNN decoder to adequately learn feature extraction. In response, we present an enhanced event-based image de-occlusion approach. Our method introduces a novel full-precision leaky integrate-and-fire (FP-LIF) mechanism to mitigate information loss within the SNN encoder. Additionally, we propose an isomorphic network knowledge distillation method to augment the feature extraction capabilities of the CNN decoder. Experimental results demonstrate the efficacy of our approach in enhancing event-camera-based SAI methodologies.

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 to the optimization of 3D bounding box regression. However, we demonstrate through experiments and mathematical proof that the 3D IoU loss suffers from abnormal gradient w.r.t. angular error and object scale, which further leads to slow convergence and suboptimal regression process, respectively. In this paper, we propose a Gradient-Corrected IoU (GCIoU) loss to achieve fast and accurate 3D bounding box regression. Specifically, a gradient correction strategy is designed to endow 3D IoU loss with a reasonable gradient. It ensures that the model converges quickly in the early stage of training, and helps to achieve fine-grained refinement of bounding boxes in the later stage. To solve suboptimal regression of 3D IoU loss for objects at different scales, we introduce a gradient rescaling strategy to adaptively optimize the step size. Finally, we integrate GCIoU Loss into multiple models to achieve stable performance gains and faster model convergence. Experiments on KITTI dataset demonstrate superiority of the proposed method. The code is available at https://github.com/ming71/GCIoU-loss.

The modern machine learning-based technologies have shown considerable potential in automatic radar scene understanding. Among these efforts, radar semantic segmentation (RSS) can provide more refined and detailed information including the moving objects and background clutters within the effective receptive field of the radar. Motivated by the success of convolutional networks in various visual computing tasks, these networks have also been introduced to solve RSS task. However, neither the regular convolution operation nor the modified ones are specific to interpret radar signals. The receptive fields of existing convolutions are defined by the object presentation in optical signals, but these two signals have different perception mechanisms. In classic radar signal processing, the object signature is detected according to a local peak response, i.e., CFAR detection. Inspired by this idea, we redefine the receptive field of the convolution operation as the peak receptive field (PRF) and propose the peak convolution operation (PeakConv) to learn the object signatures in an end-to-end network. By incorporating the proposed PeakConv layers into the encoders, our RSS network can achieve better segmentation results compared with other SoTA methods on a multi-view real-measured dataset collected from an FMCW radar. Our code for PeakConv is available at https://github.com/zlw9161/PKC.

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 the real-valued membrane potentials to 0/1 spikes to transmit information thus the multiplications of activations and weights can be replaced by additions when implemented on hardware. However, this quantization mechanism will inevitably introduce quantization error, thus causing catastrophic information loss. To address the quantization error problem, we propose a regularizing membrane potential loss (RMP-Loss) to adjust the distribution which is directly related to quantization error to a range close to the spikes. Our method is extremely simple to implement and straightforward to train an SNN. Furthermore, it is shown to consistently outperform previous state-of-the-art methods over different network architectures and datasets.

As one of the energy-efficient alternatives of conventional neural networks (CNNs), spiking neural networks (SNNs) have gained more and more interest recently. To train the deep models, some effective batch normalization (BN) techniques are proposed in SNNs. All these BNs are suggested to be used after the convolution layer as usually doing in CNNs. However, the spiking neuron is much more complex with the spatio-temporal dynamics. The regulated data flow after the BN layer will be disturbed again by the membrane potential updating operation before the firing function, i.e., the nonlinear activation. Therefore, we advocate adding another BN layer before the firing function to normalize the membrane potential again, called MPBN. To eliminate the induced time cost of MPBN, we also propose a training-inference-decoupled re-parameterization technique to fold the trained MPBN into the firing threshold. With the re-parameterization technique, the MPBN will not introduce any extra time burden in the inference. Furthermore, the MPBN can also adopt the element-wised form, while these BNs after the convolution layer can only use the channel-wised form. Experimental results show that the proposed MPBN performs well on both popular non-spiking static and neuromorphic datasets.

Learning to manipulate 3D objects in an interactive environment has been a challenging problem in Reinforcement Learning (RL). In particular, it is hard to train a policy that can generalize over objects with different semantic categories, diverse shape geometry and versatile functionality. In this study, we focused on the contact information in manipulation processes, and proposed a unified representation for critical interactions to describe different kinds of manipulation tasks. Specifically, we take advantage of the contact information generated during the RL training process and employ it as unified visual representation to predict contact map of interest. Such representation leads to an end-to-end learning framework that combined affordance based and RL based methods for the first time. Our unified framework can generalize over different types of manipulation tasks. Surprisingly, the effectiveness of such framework holds even under the multi-stage and multi-agent scenarios. We tested our method on eight types of manipulation tasks. Results showed that our methods outperform baseline algorithms, including visual affordance methods and RL methods, by a large margin on the success rate. The demonstration can be found at https://sites.google.com/view/rlafford/.

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 Spiking PointNet in the paper, the first spiking neural model for efficient deep learning on point clouds. We discover that the two huge obstacles limiting the application of SNNs in point clouds are: the intrinsic optimization obstacle of SNNs that impedes the training of a big spiking model with large time steps, and the expensive memory and computation cost of PointNet that makes training a big spiking point model unrealistic. To solve the problems simultaneously, we present a trained-less but learning-more paradigm for Spiking PointNet with theoretical justifications and in-depth experimental analysis. In specific, our Spiking PointNet is trained with only a single time step but can obtain better performance with multiple time steps inference, compared to the one trained directly with multiple time steps. We conduct various experiments on ModelNet10, ModelNet40 to demonstrate the effectiveness of Sipiking PointNet. Notably, our Spiking PointNet even can outperform its ANN counterpart, which is rare in the SNN field thus providing a potential research direction for the following work. Moreover, Spiking PointNet shows impressive speedup and storage saving in the training phase. Our code is open-sourced at https://github.com/DayongRen/Spiking-PointNet.

A challenging problem in robotics is how to control multiple robots cooperatively and safely in real-world applications. Yet, developing multi-robot control methods from the perspective of safe multi-agent reinforcement learning (MARL) has merely been studied. To fill this gap, in this study, we investigate safe MARL for multi-robot control on cooperative tasks, in which each individual robot has to not only meet its own safety constraints while maximising their reward, but also consider those of others to guarantee safe team behaviours. Firstly, we formulate the safe MARL problem as a constrained Markov game and employ policy optimisation to solve it theoretically. The proposed algorithm guarantees monotonic improvement in reward and satisfaction of safety constraints at every iteration. Secondly, as approximations to the theoretical solution, we propose two safe multi-agent policy gradient methods: Multi-Agent Constrained Policy Optimisation (MACPO) and MAPPO-Lagrangian . Thirdly, we develop the first three safe MARL benchmarks—Safe Multi-Agent MuJoCo (Safe MAMuJoCo), Safe Multi-Agent Robosuite (Safe MARobosuite) and Safe Multi-Agent Isaac Gym (Safe MAIG) to expand the toolkit of MARL and robot control research communities. Finally, experimental results on the three safe MARL benchmarks indicate that our methods can achieve state-of-the-art performance in the balance between improving reward and satisfying safety constraints compared with strong baselines. Demos and code are available at the link ( https://sites.google.com/view/aij-safe-marl/ ). 2

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 the membrane potential exceeds the firing threshold. This biomimetic mechanism of SNN appears energy-efficiency with its power sparsity and asynchronous operations on spike events. Unfortunately, with the propagation of binary spikes, the distribution of membrane potential will shift, leading to degeneration, saturation, and gradient mismatch problems, which would be disadvantageous to the network optimization and convergence. Such undesired shifts would prevent the SNN from performing well and going deep. To tackle these problems, we attempt to rectify the membrane potential distribution (MPD) by designing a novel distribution loss, MPD-Loss, which can explicitly penalize the un-desired shifts without introducing any additional operations in the inference phase. Moreover, the proposed method can also mitigate the quantization error in SNNs, which is usually ignored in other works. Experimental results demonstrate that the proposed method can directly train a deeper, larger, and better-performing SNN within fewer timesteps.

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 in the high degrees of freedom and the required cooperation among heterogeneous agents (e.g., joints of fingers). In this study, we propose the Bimanual Dexterous Hands Benchmark (Bi-DexHands), a simulator that involves two dexterous hands with tens of bimanual manipulation tasks and thousands of target objects. Tasks in Bi-DexHands are first designed to match human-level motor skills according to literature in cognitive science, and then are built in Issac Gym; this enables highly efficient RL trainings, reaching 30,000+ FPS by only one single NVIDIA RTX 3090. We provide a comprehensive benchmark for popular RL algorithms under different settings; this includes multi-agent RL, offline RL, multi-task RL, and meta RL. Our results show that PPO type on-policy algorithms can learn to solve simple manipulation tasks that are equivalent up to 48-month human baby (e.g., catching a flying object, opening a bottle), while multi-agent RL can further help to learn manipulations that require skilled bimanual cooperation (e.g., lifting a pot, stacking blocks). Despite the success on each individual task, when it comes to mastering multiple manipulation skills, existing RL algorithms fail to work in most of the multi-task and the few-shot learning tasks, which calls for more future development from the RL community. Our project is open-sourced at https://github.com/PKU-MARL/DexterousHands.

Spiking Neural Network (SNN), recognized as a type of biologically plausible architecture, has recently drawn much research attention. It transmits information by $0/1$ spikes. This bio-mimetic mechanism of SNN demonstrates extreme energy efficiency since it avoids any multiplications on neuromorphic hardware. However, the forward-passing $0/1$ spike quantization will cause information loss and accuracy degradation. To deal with this problem, the Information maximization loss (IM-Loss) that aims at maximizing the information flow in the SNN is proposed in the paper. The IM-Loss not only enhances the information expressiveness of an SNN directly but also plays a part of the role of normalization without introducing any additional operations (\textit{e.g.}, bias and scaling) in the inference phase. Additionally, we introduce a novel differentiable spike activity estimation, Evolutionary Surrogate Gradients (ESG) in SNNs. By appointing automatic evolvable surrogate gradients for spike activity function, ESG can ensure sufficient model updates at the beginning and accurate gradients at the end of the training, resulting in both easy convergence and high task performance. Experimental results on both popular non-spiking static and neuromorphic datasets show that the SNN models trained by our method outperform the current state-of-the-art algorithms.