Arrow Research search

Author name cluster

Lu Li

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

32 papers
2 author rows

Possible papers

32

EAAI Journal 2026 Journal Article

Defect detection of monocrystalline silicon wafers for photovoltaic applications using an improved you only look once version 8 small algorithm

  • Wenbo Bi
  • Xinyu Wang
  • Na Liu
  • Xu Xing
  • Lu Li
  • Hao Liu

Defects on the surface of photovoltaic monocrystalline silicon wafers, such as cracks, corners, and water stains, lead to significant performance degradation and economic losses during manufacturing. To address this, this paper proposes an improved You Only Look Once version 8 small (YOLOv8s) model. The proposed architecture integrates four strategic innovations. First, an Efficient Multi-Scale Convolution (EMSC) module is combined with the Cross-Stage Partial Bottleneck module with two convolutions (C2f) to enhance multi-scale feature extraction capabilities. Second, Spatial Pyramid Pooling-Fast (SPPF) is fused with the Large Separable Kernel Attention (LSKA) module to overcome limitations in processing local details. Third, the Dysample dynamic upsampling operator is introduced to maintain a compact model size while effectively improving detection speed. Finally, the Normalized Wasserstein Distance (NWD) is utilized as the loss function to address the sensitivity of the Intersection over Union (IoU) metric to positional deviations, enhancing precision for small targets. Experimental results demonstrate that the Efficient Lightweight Detection Network (ELDN) achieves superior performance on the validation set with a mean Average Precision (mAP) of 92. 8%. Notably, it exhibits robust generalization on an independent external test set, attaining a mAP of 92. 6%. Validation confirms that YOLOv8s-ELDN consistently outperforms mainstream models. Future research will focus on further optimizing efficiency for deployment on resource-constrained edge devices and addressing defect detection in complex manufacturing environments.

Details DOI

EAAI Journal 2025 Journal Article

A framework for super-resolution of side-scan sonar images: Combination of variational Bayes and regional feature selection

  • Xin Wen
  • Chensheng Cheng
  • Lu Li
  • Feihu Zhang
  • Guang Pan

Side-scan sonar is widely used in ocean exploration due to its broad search range and strong identification capabilities. However, the inherent characteristics of acoustic images often result in poor image quality, negatively impacting subsequent downstream tasks’ accuracy. Image super-resolution (SR) technology based on deep learning technology is employed to address this issue. Despite this, existing SR models face two main challenges when applied to side-scan sonar images: (1) less data in side-scan sonar images causes the model overfitting problem; (2) less effective features in side-scan sonar images cause lower efficiency. To overcome these challenges, this paper proposes a deep learning framework that integrates a Bayesian structure with region-based feature selection. First, we introduce a rolling region selection method to extract key features of interest from side-scan sonar images, enhancing efficiency without compromising quality. Additionally, we replace traditional Convolutional Neural Networks (CNN) with Variational Bayes Convolutional Neural Networks (VB-CNN) to perform the SR task, improving generalization on small datasets and mitigating the risk of overfitting. Experiments conducted on the Side-Scan Sonar Visual Object Classes (SSS-VOC) dataset and other datasets demonstrate our proposed approach’s effectiveness through both qualitative and quantitative comparisons.

Details DOI

ICML Conference 2025 Conference Paper

AI for Global Climate Cooperation: Modeling Global Climate Negotiations, Agreements, and Long-Term Cooperation in RICE-N

  • Tianyu Zhang
  • Andrew Robert Williams
  • Phillip Wozny
  • Kai-Hendrik Cohrs
  • Koen Ponse
  • Marco Jiralerspong
  • Soham R. Phade
  • Sunil Srinivasa

Global cooperation on climate change mitigation is essential to limit temperature increases while supporting long-term, equitable economic growth and sustainable development. Achieving such cooperation among diverse regions, each with different incentives, in a dynamic environment shaped by complex geopolitical and economic factors, without a central authority, is a profoundly challenging game-theoretic problem. This article introduces RICE-N, a multi-region integrated assessment model that simulates the global climate, economy, and climate negotiations and agreements. RICE-N uses multi-agent reinforcement learning (MARL) to encourage agents to develop strategic behaviors based on the environmental dynamics and the actions of the others. We present two negotiation protocols: (1) Bilateral Negotiation, an exemplary protocol and (2) Basic Club, inspired from Climate Clubs and the carbon border adjustment mechanism (Nordhaus, 2015; Comissions, 2022). We compare their impact against a no-negotiation baseline with various mitigation strategies, showing that both protocols significantly reduce temperature growth at the cost of a minor drop in production while ensuring a more equitable distribution of the emission reduction costs.

Details

IROS Conference 2025 Conference Paper

Bag-of-Word-Groups (BoWG): A Robust and Efficient Loop Closure Detection Method Under Perceptual Aliasing

  • Xiang Fei
  • Tina Tian
  • Howie Choset
  • Lu Li

Loop closure is critical in Simultaneous Localization and Mapping (SLAM) systems to reduce accumulative drift and ensure global mapping consistency. However, conventional methods struggle in perceptually aliased environments, such as narrow pipes, due to vector quantization, feature sparsity, and repetitive textures, while existing solutions often incur high computational costs. This paper presents Bag-of-Word-Groups (BoWG), a novel loop closure detection method that achieves superior precision-recall, robustness, and computational efficiency. The core innovation lies in the introduction of word groups, which captures the spatial co-occurrence and proximity of visual words to construct an online dictionary. Additionally, drawing inspiration from probabilistic transition models, we incorporate temporal consistency directly into similarity computation with an adaptive scheme, substantially improving precision-recall performance. The method is further strengthened by a feature distribution analysis module and dedicated post-verification mechanisms. To evaluate the effectiveness of our method, we conduct experiments on both public datasets and a confined-pipe dataset we constructed. Results demonstrate that BoWG surpasses state-of-the-art methods—including both traditional and learning-based approaches—in terms of precision-recall and computational efficiency. Our approach also exhibits excellent scalability, achieving an average processing time of 16 ms per image across 17, 565 images in the Bicocca25b dataset. The source code is available at: https://github.com/EdgarFx/BoWG.

Details

ICRA Conference 2025 Conference Paper

Bio-Inspired Distributed Neural Locomotion Controller (D-NLC) for Robust Locomotion and Emergent Behaviors

  • Zhikai Zhang
  • Siqi Guo
  • Henry Kou
  • Ishayu Shikhare
  • Howie Choset
  • Lu Li

Despite having fewer neurons than more complex life forms, insects are still capable of producing astonishing locomotive behaviors, such as traversing diverse environments and making rapid gait adaptations after extreme injury or autotomy. Biologists attribute this to a chain of segmental neuron clusters (ganglia) within insect nervous systems, which act as distributed self-organizing sensorimotor control units. Inspired by the neural structure of the Carausius morosus, the common stick insect, this work introduces the Distributed Neural Locomotion Controller (D-NLC), a modular control framework that utilizes local proprioceptive feedback to modulate joint-level Central Pattern Generator (CPG) signals to produce emergent locomotive behaviors. This framework was implemented on a modular legged robot with distributed jointlevel embedded computing units. In addition, assessments were conducted on the framework's performance and behavior in various experimental settings. Based on real-world experiments, we observe an overall 31. 3% average increase in curvilinear motion performance under external (terrain) and internal (amputation) perturbation compared to a centralized predefined gait controller. This difference is statistically significant $(P \ll 0. 05)$ for larger perturbations but not for single-leg amputations. Experiments with perturbation-induced leg stance duration and leg phase-difference analysis further validated our hypothesis regarding D-NLC's role in the robust perceptive locomotion and self-emergent gait adaptation against complex unforeseen perturbations. This proposed control framework does not require any numerical optimization or weight training processes, which are time-consuming and computationally expensive. To the best of our knowledge, this framework is the first bio-inspired neural controller deployed on a distributed embedded system. More info at https://eigenbot-dnlc.github.io.

Details

ECAI Conference 2025 Conference Paper

Deep Learning and Explainable AI: New Pathways to Genetic Insights

  • Chenyu Wang
  • Chaoying Zuo
  • Zihan Su
  • Yuhang Xing
  • Lu Li
  • Maojun Wang
  • Zeyu Zhang 0004

Deep learning-based AI models have been extensively applied in genomics, achieving remarkable success across diverse applications. As these models gain prominence, there exists an urgent need for interpretability methods to establish trustworthiness in model-driven decisions. For genetic researchers, interpretable insights derived from these models hold significant value in providing novel perspectives for understanding biological processes. Current interpretability analyses in genomics predominantly rely on intuition and experience rather than rigorous theoretical foundations. In this review, we categorize interpretability methods into input-based and model-based approaches, while critically evaluating their limitations through concrete biological application scenarios. Furthermore, we establish theoretical underpinnings to elucidate the origins of these constraints through formal mathematical demonstrations, aiming to assist genetic researchers in better understanding and designing models in the future. Finally, we provide feasible suggestions for future research on interpretability in the field of genetics.

Details

IJCAI Conference 2025 Conference Paper

ListenNet: A Lightweight Spatio-Temporal Enhancement Nested Network for Auditory Attention Detection

  • Cunhang Fan
  • Xiaoke Yang
  • Hongyu Zhang
  • Ying Chen
  • Lu Li
  • Jian Zhou
  • Zhao Lv

Auditory attention detection (AAD) aims to identify the direction of the attended speaker in multi-speaker environments from brain signals, such as Electroencephalography (EEG) signals. However, existing EEG-based AAD methods overlook the spatio-temporal dependencies of EEG signals, limiting their decoding and generalization abilities. To address these issues, this paper proposes a Lightweight Spatio-Temporal Enhancement Nested Network (ListenNet) for AAD. The ListenNet has three key components: Spatio-temporal Dependency Encoder (STDE), Multi-scale Temporal Enhancement (MSTE), and Cross-Nested Attention (CNA). The STDE reconstructs dependencies between consecutive time windows across channels, improving the robustness of dynamic pattern extraction. The MSTE captures temporal features at multiple scales to represent both fine-grained and long-range temporal patterns. In addition, the CNA integrates hierarchical features more effectively through novel dynamic attention mechanisms to capture deep spatio-temporal correlations. Experimental results on three public datasets demonstrate the superiority of ListenNet over state-of-the-art methods in both subject-dependent and challenging subject-independent settings, while reducing the trainable parameter count by approximately 7 times. Code is available at: https: //github. com/fchest/ListenNet.

PDF Details DOI

ICLR Conference 2025 Conference Paper

MAP: Low-compute Model Merging with Amortized Pareto Fronts via Quadratic Approximation

  • Lu Li
  • Tianyu Zhang
  • Zhiqi Bu
  • Suyuchen Wang
  • Huan He
  • Jie Fu 0001
  • Yonghui Wu 0001
  • Jiang Bian 0002

Model merging has emerged as an effective approach to combining multiple single-task models into a multitask model. This process typically involves computing a weighted average of the model parameters without additional training. Existing model-merging methods focus on improving average task accuracy. However, interference and conflicts between the objectives of different tasks can lead to trade-offs during the merging process. In real-world applications, a set of solutions with various trade-offs can be more informative, helping practitioners make decisions based on diverse preferences. In this paper, we introduce a novel and low-compute algorithm, Model Merging with Amortized Pareto Front (MAP). MAP efficiently identifies a Pareto set of scaling coefficients for merging multiple models, reflecting the trade-offs involved. It amortizes the substantial computational cost of evaluations needed to estimate the Pareto front by using quadratic approximation surrogate models derived from a preselected set of scaling coefficients. Experimental results on vision and natural language processing tasks demonstrate that MAP can accurately identify the Pareto front, providing practitioners with flexible solutions to balance competing task objectives. We also introduce Bayesian MAP for scenarios with a relatively low number of tasks and Nested MAP for situations with a high number of tasks, further reducing the computational cost of evaluation.

Details

IJCAI Conference 2025 Conference Paper

MHANet: Multi-scale Hybrid Attention Network for Auditory Attention Detection

  • Lu Li
  • Cunhang Fan
  • Hongyu Zhang
  • Jingjing Zhang
  • Xiaoke Yang
  • Jian Zhou
  • Zhao Lv

Auditory attention detection (AAD) aims to detect the target speaker in a multi-talker environment from brain signals, such as electroencephalography (EEG), which has made great progress. However, most AAD methods solely utilize attention mechanisms sequentially and overlook valuable multi-scale contextual information within EEG signals, limiting their ability to capture long-short range spatiotemporal dependencies simultaneously. To address these issues, this paper proposes a multi-scale hybrid attention network (MHANet) for AAD, which consists of the multi-scale hybrid attention (MHA) module and the spatiotemporal convolution (STC) module. Specifically, MHA combines channel attention and multi-scale temporal and global attention mechanisms. This effectively extracts multi-scale temporal patterns within EEG signals and captures long-short range spatiotemporal dependencies simultaneously. To further improve the performance of AAD, STC utilizes temporal and spatial convolutions to aggregate expressive spatiotemporal representations. Experimental results show that the proposed MHANet achieves state-of-the-art performance with fewer trainable parameters across three datasets, 3 times lower than that of the most advanced model. Code is available at: https: //github. com/fchest/MHANet.

PDF Details DOI

ICML Conference 2025 Conference Paper

Network Sparsity Unlocks the Scaling Potential of Deep Reinforcement Learning

  • Guozheng Ma
  • Lu Li
  • Zilin Wang 0002
  • Li Shen 0008
  • Pierre-Luc Bacon
  • Dacheng Tao

Effectively scaling up deep reinforcement learning models has proven notoriously difficult due to network pathologies during training, motivating various targeted interventions such as periodic reset and architectural advances such as layer normalization. Instead of pursuing more complex modifications, we show that introducing static network sparsity alone can unlock further scaling potential beyond their dense counterparts with state-of-the-art architectures. This is achieved through simple one-shot random pruning, where a predetermined percentage of network weights are randomly removed once before training. Our analysis reveals that, in contrast to naively scaling up dense DRL networks, such sparse networks achieve both higher parameter efficiency for network expressivity and stronger resistance to optimization challenges like plasticity loss and gradient interference. We further extend our evaluation to visual and streaming RL scenarios, demonstrating the consistent benefits of network sparsity.

Details

AAAI Conference 2025 Conference Paper

Self-Explainable Graph Transformer for Link Sign Prediction

  • Lu Li
  • Jiale Liu
  • Xingyu Ji
  • Maojun Wang
  • Zeyu Zhang

Signed Graph Neural Networks (SGNNs) have been shown to be effective in analyzing complex patterns in real-world situations where positive and negative links coexist. However, SGNN models suffer from poor explainability, which limit their adoptions in critical scenarios that require understanding the rationale behind predictions. To the best of our knowledge, there is currently no research work on the explainability of the SGNN models. Our goal is to address the explainability of decision-making for the downstream task of link sign prediction specific to signed graph neural networks. Since post-hoc explanations are not derived directly from the models, they may be biased and misrepresent the true explanations. Therefore, in this paper we introduce a Self-Explainable Signed Graph transformer (SE-SGformer) framework, which can not only outputs explainable information while ensuring high prediction accuracy. Specifically, we propose a new Transformer architecture for signed graphs and theoretically demonstrate that using positional encoding based on signed random walks has greater expressive power than current SGNN methods and other positional encoding graph Transformer-based approaches. We construct a novel explainable decision process by discovering the K-nearest (farthest) positive (negative) neighbors of a node to replace the neural network-based decoder for predicting edge signs. These K positive (negative) neighbors represent crucial information about the formation of positive (negative) edges between nodes and thus can serve as important explanatory information in the decision-making process. We conducted experiments on several real-world datasets to validate the effectiveness of SE-SGformer, which outperforms the state-of-the-art methods by improving 2.2% prediction accuracy and 73.1% explainablity accuracy in the best-case scenario.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Stable Gradients for Stable Learning at Scale in Deep Reinforcement Learning

  • Roger Creus Castanyer
  • Johan Obando Ceron
  • Lu Li
  • Pierre-Luc Bacon
  • Glen Berseth
  • Aaron Courville
  • Pablo Samuel Castro

Scaling deep reinforcement learning networks is challenging and often results in degraded performance, yet the root causes of this failure mode remain poorly understood. Several recent works have proposed mechanisms to address this, but they are often complex and fail to highlight the causes underlying this difficulty. In this work, we conduct a series of empirical analyses which suggest that the combination of non-stationarity with gradient pathologies, due to suboptimal architectural choices, underlie the challenges of scale. We propose a series of direct interventions that stabilize gradient flow, enabling robust performance across a range of network depths and widths. Our interventions are simple to implement and compatible with well-established algorithms, and result in an effective mechanism that enables strong performance even at large scales. We validate our findings on a variety of agents and suites of environments.

PDF Details

ICLR Conference 2025 Conference Paper

VCR: A Task for Pixel-Level Complex Reasoning in Vision Language Models via Restoring Occluded Text

  • Tianyu Zhang
  • Suyuchen Wang
  • Lu Li
  • Ge Zhang 0009
  • Perouz Taslakian
  • Sai Rajeswar
  • Jie Fu 0001
  • Bang Liu 0003

We introduce Visual Caption Restoration (VCR), a novel vision-language task that challenges models to accurately restore partially obscured texts using pixel-level hints within images through complex reasoning. This task stems from the observation that text embedded in images intrinsically differs from common visual elements and text due to the need to align the modalities of vision, text, and text embedded in images. While many works incorporate text into images for visual question answering, they mostly rely on OCR or masked language modeling, reducing the task to text-based processing. However, text-based processing becomes ineffective in VCR as accurate text restoration depends on the combined information from provided images, context, and subtle cues from the tiny, exposed areas of masked texts. We develop a pipeline to generate synthetic images for the VCR task using image-caption pairs, with adjustable caption visibility to control the task difficulty. With this pipeline, we construct VCR-WIKI for VCR using Wikipedia images with captions, including 2.11M English and 346K Chinese training entities, plus 5K validation and 5K test entities in both languages, each in easy and hard configurations. We also make a hidden test set, VCR-HIDDEN, to avoid potential overfitting on VCR-WIKI. Our results reveal that current vision-language models significantly lag behind human performance in the VCR task, and merely fine-tuning the models on our dataset does not lead to notable improvements. We release VCR-WIKI and the data construction code to facilitate future research.

Details

NeurIPS Conference 2024 Conference Paper

DropEdge not Foolproof: Effective Augmentation Method for Signed Graph Neural Networks

  • Zeyu Zhang
  • Lu Li
  • Shuyan Wan
  • Sijie Wang
  • Zhiyi Wang
  • Zhiyuan Lu
  • Dong Hao
  • Wanli Li

Signed graphs can model friendly or antagonistic relations where edges are annotated with a positive or negative sign. The main downstream task in signed graph analysis is $\textit{link sign prediction}$. Signed Graph Neural Networks (SGNNs) have been widely used for signed graph representation learning. While significant progress has been made in SGNNs research, two issues (i. e. , graph sparsity and unbalanced triangles) persist in the current SGNN models. We aim to alleviate these issues through data augmentation ($\textit{DA}$) techniques which have demonstrated effectiveness in improving the performance of graph neural networks. However, most graph augmentation methods are primarily aimed at graph-level and node-level tasks (e. g. , graph classification and node classification) and cannot be directly applied to signed graphs due to the lack of side information (e. g. , node features and label information) in available real-world signed graph datasets. Random $\textit{DropEdge} $is one of the few $\textit{DA}$ methods that can be directly used for signed graph data augmentation, but its effectiveness is still unknown. In this paper, we first provide the generalization bound for the SGNN model and demonstrate from both experimental and theoretical perspectives that the random $\textit{DropEdge}$ cannot improve the performance of link sign prediction. Therefore, we propose a novel signed graph augmentation method, $\underline{S}$igned $\underline{G}$raph $\underline{A}$ugmentation framework (SGA). Specifically, SGA first integrates a structure augmentation module to detect candidate edges solely based on network information. Furthermore, SGA incorporates a novel strategy to select beneficial candidates. Finally, SGA introduces a novel data augmentation perspective to enhance the training process of SGNNs. Experiment results on six real-world datasets demonstrate that SGA effectively boosts the performance of diverse SGNN models, achieving improvements of up to 32. 3\% in F1-micro for SGCN on the Slashdot dataset in the link sign prediction task.

PDF Details DOI

EAAI Journal 2024 Journal Article

Ensemble CART surrogate-assisted automatic multi-objective rough fuzzy clustering algorithm for unsupervised image segmentation

  • Feng Zhao
  • Zihan Tang
  • Zhilei Xiao
  • Hanqiang Liu
  • Jiulun Fan
  • Lu Li

Multi-objective clustering algorithms (MOCAs) are popular in unsupervised image segmentation due to their merit of meeting multiple segmentation requirements and the prospect of automatically estimating the number of clusters. However, most of them suffer from high time costs and are easily to be influenced by the uncertainty when handling real complex images. To address these issues, we propose an ensemble classification and regression tree (CART) surrogate-assisted automatic multi-objective rough fuzzy clustering (ECS-AMRFC) algorithm for unsupervised image segmentation. Firstly, a cluster medoid-based encoding scheme is employed to represent solutions with different number of clusters and meanwhile lessen the length of encoding. Then, we design an ensemble CART as the surrogate model to significantly reduce the computational burden. Moreover, a surrogate model management strategy is proposed to accelerate the optimization and enhance the quality of surrogate modeling. To handle the uncertainty in data, we extend the rough fuzzy clustering into MOCAs and construct three complementary objective functions to seek proper cluster medoids from multiple perspectives. In addition, the Gaussian kernel is introduced into the objective functions to handle image pixels that cannot separate linearly in the feature space. Finally, a kernelized rough fuzzy clustering validity index is defined to automatically select the optimal solution with no requirements of any prior knowledge. Experiments show that ECS-AMRFC not only identifies appropriate number of clusters on different kinds of images, but also obtains better segmentation results than state-of-the-art rough fuzzy clustering algorithms and automatic MOCAs.

Details DOI

AAAI Conference 2024 Conference Paper

Explore 3D Dance Generation via Reward Model from Automatically-Ranked Demonstrations

  • Zilin Wang
  • Haolin Zhuang
  • Lu Li
  • Yinmin Zhang
  • Junjie Zhong
  • Jun Chen
  • Yu Yang
  • Boshi Tang

This paper presents an Exploratory 3D Dance generation framework, E3D2, designed to address the exploration capability deficiency in existing music-conditioned 3D dance generation models. Current models often generate monotonous and simplistic dance sequences that misalign with human preferences because they lack exploration capabilities.The E3D2 framework involves a reward model trained from automatically-ranked dance demonstrations, which then guides the reinforcement learning process. This approach encourages the agent to explore and generate high quality and diverse dance movement sequences. The soundness of the reward model is both theoretically and experimentally validated. Empirical experiments demonstrate the effectiveness of E3D2 on the AIST++ dataset.

PDF Details DOI

ICLR Conference 2024 Conference Paper

Mathematical Justification of Hard Negative Mining via Isometric Approximation Theorem

  • Albert Xu
  • Jhih-Yi Hsieh
  • Bhaskar Vundurthy
  • Nithya Kemp
  • Eliana Cohen
  • Lu Li
  • Howie Choset

In deep metric learning, the triplet loss has emerged as a popular method to learn many computer vision and natural language processing tasks such as facial recognition, object detection, and visual-semantic embeddings. One issue that plagues the triplet loss is network collapse, an undesirable phenomenon where the network projects the embeddings of all data onto a single point. Researchers predominately solve this problem by using triplet mining strategies. While hard negative mining is the most effective of these strategies, existing formulations lack strong theoretical justification for their empirical success. In this paper, we utilize the mathematical theory of isometric approximation to show an equivalence between the triplet loss sampled by hard negative mining and an optimization problem that minimizes a Hausdorff-like distance between the neural network and its ideal counterpart function. This provides the theoretical justifications for hard negative mining's empirical efficacy. Experiments performed on the Market-1501 and Stanford Online Products datasets with various network architectures corroborate our theoretical findings, indicating that network collapse tends to happen when batch size is too large or embedding dimension is too small. In addition, our novel application of the isometric approximation theorem provides the groundwork for future forms of hard negative mining that avoid network collapse.

Details

AAMAS Conference 2024 Conference Paper

Normalization Enhances Generalization in Visual Reinforcement Learning

  • Lu Li
  • Jiafei Lyu
  • Guozheng Ma
  • Zilin Wang
  • Zhenjie Yang
  • Xiu Li
  • Zhiheng Li

Recent advances in visual reinforcement learning (RL) have led to impressive success in handling complex tasks. However, these methods have demonstrated limited generalization capability to visual disturbances, which poses a significant challenge to their real-world application and adaptability. Though normalization techniques have demonstrated huge success in supervised and unsupervised learning, their applications in visual RL are still scarce. In this paper, we explore the potential benefits of integrating normalization into visual RL methods with respect to generalization performance. We find that, perhaps surprisingly, incorporating suitable normalization techniques is sufficient to enhance the generalization capabilities, without any additional special design. We utilize the combination of two normalization techniques, CrossNorm and SelfNorm, for generalizable visual RL. Extensive experiments are conducted on DMControl Generalization Benchmark, CARLA, and ProcGen Benchmark to validate the effectiveness of our method. We show that our method significantly improves generalization capability while only marginally affecting sample efficiency. In particular, when integrated with DrQ-v2, our method enhances the test performance of DrQ-v2 on CARLA across various scenarios, from 14% of the training performance to 97%. Our project page: https: //sites. google. com/view/norm-generalization-vrl/home

PDF

ICLR Conference 2024 Conference Paper

Revisiting Plasticity in Visual Reinforcement Learning: Data, Modules and Training Stages

  • Guozheng Ma
  • Lu Li
  • Sen Zhang 0006
  • Zixuan Liu 0002
  • Zhen Wang 0030
  • Yixin Chen 0001
  • Li Shen 0008
  • Xueqian Wang 0001

Plasticity, the ability of a neural network to evolve with new data, is crucial for high-performance and sample-efficient visual reinforcement learning (VRL). Although methods like resetting and regularization can potentially mitigate plasticity loss, the influences of various components within the VRL framework on the agent's plasticity are still poorly understood. In this work, we conduct a systematic empirical exploration focusing on three primary underexplored facets and derive the following insightful conclusions: (1) data augmentation is essential in maintaining plasticity; (2) the critic's plasticity loss serves as the principal bottleneck impeding efficient training; and (3) without timely intervention to recover critic's plasticity in the early stages, its loss becomes catastrophic. These insights suggest a novel strategy to address the high replay ratio (RR) dilemma, where exacerbated plasticity loss hinders the potential improvements of sample efficiency brought by increased reuse frequency. Rather than setting a static RR for the entire training process, we propose Adaptive RR, which dynamically adjusts the RR based on the critic’s plasticity level. Extensive evaluations indicate that Adaptive RR not only avoids catastrophic plasticity loss in the early stages but also benefits from more frequent reuse in later phases, resulting in superior sample efficiency.

Details

NeurIPS Conference 2023 Conference Paper

Learning Better with Less: Effective Augmentation for Sample-Efficient Visual Reinforcement Learning

  • Guozheng Ma
  • Linrui Zhang
  • Haoyu Wang
  • Lu Li
  • Zilin Wang
  • Zhen Wang
  • Li Shen
  • Xueqian Wang

Data augmentation (DA) is a crucial technique for enhancing the sample efficiency of visual reinforcement learning (RL) algorithms. Notably, employing simple observation transformations alone can yield outstanding performance without extra auxiliary representation tasks or pre-trained encoders. However, it remains unclear which attributes of DA account for its effectiveness in achieving sample-efficient visual RL. To investigate this issue and further explore the potential of DA, this work conducts comprehensive experiments to assess the impact of DA's attributes on its efficacy and provides the following insights and improvements: (1) For individual DA operations, we reveal that both ample spatial diversity and slight hardness are indispensable. Building on this finding, we introduce Random PadResize (Rand PR), a new DA operation that offers abundant spatial diversity with minimal hardness. (2) For multi-type DA fusion schemes, the increased DA hardness and unstable data distribution result in the current fusion schemes being unable to achieve higher sample efficiency than their corresponding individual operations. Taking the non-stationary nature of RL into account, we propose a RL-tailored multi-type DA fusion scheme called Cycling Augmentation (CycAug), which performs periodic cycles of different DA operations to increase type diversity while maintaining data distribution consistency. Extensive evaluations on the DeepMind Control suite and CARLA driving simulator demonstrate that our methods achieve superior sample efficiency compared with the prior state-of-the-art methods.

PDF Details

AAAI Conference 2023 Conference Paper

READ: Large-Scale Neural Scene Rendering for Autonomous Driving

  • Zhuopeng Li
  • Lu Li
  • Jianke Zhu

With the development of advanced driver assistance systems~(ADAS) and autonomous vehicles, conducting experiments in various scenarios becomes an urgent need. Although having been capable of synthesizing photo-realistic street scenes, conventional image-to-image translation methods cannot produce coherent scenes due to the lack of 3D information. In this paper, a large-scale neural rendering method is proposed to synthesize the autonomous driving scene~(READ), which makes it possible to generate large-scale driving scenes in real time on a PC through a variety of sampling schemes. In order to effectively represent driving scenarios, we propose an ω-net rendering network to learn neural descriptors from sparse point clouds. Our model can not only synthesize photo-realistic driving scenes but also stitch and edit them. The promising experimental results show that our model performs well in large-scale driving scenarios.

PDF Details DOI

IROS Conference 2023 Conference Paper

Real-Time Video Inpainting for RGB-D Pipeline Reconstruction

  • Luyuan Wang
  • Tina Tian
  • Xinzhi Yan
  • Fujun Ruan
  • G. Jaya Aadityaa
  • Howie Choset
  • Lu Li

This paper presents a Video Inpainting algorithm that enables monocular-camera-laser-based pipeline inspection robots to capture both color and 3D information using only one video stream. Conventional monocular-camera-laser inspection methods are limited to capture either 2D color images or 3D point clouds since the laser tends to overexpose the actual color of the scanning area. We propose a real-time Video Inpainting method to solve this problem with minimal hardware needs that can be easily integrated with conventional pipeline profiling robots. The algorithm is accelerated by two components: a lightweight network that directly predicts the complete optical flow and simplifies the algorithm pipeline, and the Polar coordinate transformation, which significantly reduces the image processing compexity. Real-world experiments demonstrate that our online algorithm has comparable or better color estimation accuracy against state-of-the-art offline algorithms, while is capable of running at 23 frames per second (FPS) on a laptop computer with a resolution of 1024 × 1024 pixels. In addition, we verify that this method can be used for video pre-processing for downstream tasks that require high-quality visual inputs, such as Simultaneously Localization and Mapping (SLAM). To the best of our knowledge, this is the first real-time Video Inpainting algorithm that can be used for in-pipe environments, serving as an important building block for highly compact RGB-D inspection sensors and robots for the pipeline industry.

Details

IROS Conference 2023 Conference Paper

Toward Closed-Loop Additive Manufacturing: Paradigm Shift in Fabrication, Inspection, and Repair

  • Manpreet Singh
  • Fujun Ruan
  • Albert Xu
  • Yuchen Wu
  • Archit Rungta
  • Luyuan Wang
  • Kevin Song
  • Howie Choset

Increased usage of additive manufacturing (AM) in various industries has solidified its role as an advanced manufacturing technique. However, there is an inherent lack of reliability in AM processes, particularly common in extrusion or deposition-based methods due to the stochastic nature of ma-terial deposition. This necessitates an intelligent manufacturing solution to address the drawbacks of AM. Thus, we propose a novel layer-wise approach toward closed-loop AM, which is capable of in-situ monitoring and repairing geometric defects. In this paper, we present a system that uses a robotic AM experimental platform that mimics a conventional open-loop fabrication setup, which we augment into a closed-loop system using two add-ons: in-situ inspection subsystem and online process correction subsystem. The in-situ inspection subsystem collects 3D point cloud scans and compares them against a reference CAD model, categorizing geometric deviations as positive or negative defects. Then the subsequent online process correction subsystem uses a re-plan and/or repair strategy to address the positive and/or negative defects, respectively. To evaluate this idea, we conducted three experiments on parts with manually induced defects to investigate the system's ability to repair those parts, thereby reducing defects, improving part accuracy, and enhancing mechanical properties. Comparing the defective and repaired parts, we observe a reduction in defect percent by volume from 10. 7% to 1. 3%, an improvement in geometric tolerance from 3. 86% error to 0. 08% error, and an increase in the part's breaking load from 4. 77 kN to 6. 31 kN. These experiments prove that our layer-wise closed-loop additive manufacturing approach improves the quality, tolerance, and reliability of plastic 3D printed parts, with the potential to extend to other extrusion/deposition-based AM processes, or even subtractive manufacturing and hybrid manufacturing methods.

Details

IROS Conference 2023 Conference Paper

Visual-Inertial-Laser-Lidar (VILL) SLAM: Real-Time Dense RGB-D Mapping for Pipe Environments

  • Tina Tian
  • Luyuan Wang
  • Xinzhi Yan
  • Fujun Ruan
  • G. Jaya Aadityaa
  • Howie Choset
  • Lu Li

Robotic solutions for pipeline inspection promise enhancement of human labor by automating data acquisition for pipe condition assessments, which are vital for the early detection of pipe anomalies and the prevention of hazardous leakages and explosions. Through simultaneous localization and mapping (SLAM), colorized 3D reconstructions of the pipe's inner surface can be generated, providing a more comprehensive digital record of the pipes compared to conventional vision-only inspection. Designed for generic environments, most SLAM methods suffer limited accuracy and substantial accumulative drift in confined and featureless spaces such as pipelines, due to a lack of suitable sensor hardware and state estimation techniques. In this research, we present VILL-SLAM: a dense RGB-D SLAM algorithm that combines a monocular camera (V), an inertial sensor (I), a ring-shaped laser profiler (L), and a Lidar (L) into a compact sensor package optimized for in-pipe operations. By fusing complementary visual and depth information from the color camera, laser profiling, and Lidar measurement, our method overcomes the challenges of metric scale mapping in conventional SLAM methods, despite its monocular configuration. To further improve localization accuracy, we utilize the pipe geometry to formulate two unique optimization factors that effectively constrain odometer drift. To validate our method, we conducted real-world experiments in physical pipes, comparing the performance of our approach against other state-of-the-art algorithms. The proposed SLAM framework achieved 6. 6 times drift improvement with 0. 84% mean odometry drift over 22 meters and a mean pointwise 3D scanning error of 0. 88mm in 12-inch diameter pipes. This research represents a significant advancement in miniature in-pipe inspection, localization, and mapping sensing techniques. It has the potential to become a core enabling technology for the next generation of highly capable in-pipe robots, capable of reconstructing photo-realistic 3D pipe scans and providing disruptive pipe locating and georeferencing capabilities.

Details

ICRA Conference 2022 Conference Paper

Design of a Biomimetic Tactile Sensor for Material Classification

  • Kevin Dai
  • Xinyu Wang
  • Allison M. Rojas
  • Evan Harber
  • Yu Tian
  • Nicholas Paiva
  • Joseph Gnehm
  • Evan Schindewolf

Tactile sensing typically involves active exploration of unknown surfaces and objects, making it especially effective at processing the characteristics of materials and textures. A key property extracted by human tactile perception in material classification is surface roughness, which relies on measuring vibratory signals using the multi-layered fingertip structure. Existing robotic systems lack tactile sensors that are able to provide high dynamic sensing ranges, perceive material properties, and maintain a low hardware cost. In this work, we introduce the reference design and fabrication procedure of a miniature and low-cost tactile sensor consisting of a biomimetic cutaneous structure, including the artificial fingerprint, dermis, epidermis, and an embedded magnet-sensor structure which serves as a mechanoreceptor for converting mechanical information to digital signals. The presented sensor is capable of detecting high-resolution magnetic field data through the Hall effect and creating high-dimensional time-frequency domain features for material texture classification. Additionally, we investigate the effects of different superficial sensor fingerprint patterns for classifying materials through both simulation and physical experimentation. After extracting time series and frequency domain features, we assess a k-nearest neighbors classifier for distinguishing between different materials. The results from our experiments show that our biomimetic tactile sensors with fingerprint ridges can classify materials with more than 7. 7% higher accuracy and lower variability than ridge-less sensors. These results, along with the low cost and customizability of our sensor, demonstrate high potential for lowering the barrier to entry for a wide array of robotic applications, including modelless tactile sensing for texture classification, material inspection, and object recognition.

Details

JBHI Journal 2021 Journal Article

Deep Learning for Hemorrhagic Lesion Detection and Segmentation on Brain CT Images

  • Lu Li
  • Meng Wei
  • Bo Liu
  • Kunakorn Atchaneeyasakul
  • Fugen Zhou
  • Zehao Pan
  • Shimran A. Kumar
  • Jason Y. Zhang

Stroke is an acute cerebral vascular disease that is likely to cause long-term disabilities and death. Immediate emergency care with accurate diagnosis of computed tomographic (CT) images is crucial for dealing with a hemorrhagic stroke. However, due to the high variability of a stroke's location, contrast, and shape, it is challenging and time-consuming even for experienced radiologists to locate them. In this paper, we propose a U-net based deep learning framework to automatically detect and segment hemorrhage strokes in CT brain images. The input of the network is built by concatenating the flipped image with the original CT slice which introduces symmetry constraints of the brain images into the proposed model. This enhances the contrast between hemorrhagic area and normal brain tissue. Various Deep Learning topologies are compared by varying the layers, batch normalization, dilation rates, and pre-train models. This could increase the respective filed and preserves more information on lesion characteristics. Besides, the adversarial training is also adopted in the proposed network to improve the accuracy of the segmentation. The proposed model is trained and evaluated on two different datasets, which achieve the competitive performance with human experts with the highest location accuracy 0. 9859 for detection, 0. 8033 Dice score, and 0. 6919 IoU for segmentation. The results demonstrate the effectiveness, robustness, and advantages of the proposed deep learning model in automatically hemorrhage lesion diagnosis, which make it possible to be a clinical decision support tool in stroke diagnosis.

Details DOI

ICRA Conference 2021 Conference Paper

Robotic Cardinal Vein Microinjection of Zebrafish Larvae Based on 3D Positioning

  • Mingzhu Sun
  • Lu Li
  • Yatong Yao
  • Yiwen Wang
  • Huiying Gong
  • Qian Gao
  • Dongyan Chen
  • Xin Zhao 0010

Zebrafish (Danio Rerio) larvae have long been an important model organism for biomedicine and drug discovery. It is difficult to deliver the external materials into the circulatory system by conventional exposing administration, while vein microinjection is more efficient but more challenging. In this paper, a robotic cardinal vein microinjection system was presented for zebrafish larvae. The key points of injection pipette during penetration were analyzed, and their 3D positions were identified online by combining motion analysis and visual calibration. 3D path planning of the pipette tip was designed automatedly before injection. The injection system was evaluated by injecting green fluorescent microspheres into 20 zebrafish larvae. The experiment results show that the fluorescent microspheres spread rapidly in the blood circulation system after injection, achieving a success rate of 85. 0% and a survival rate of 94. 1%. The superior performance provided by the system will significantly facilitate tumor xenografts and canner studies on zebrafish.

Details

ICRA Conference 2021 Conference Paper

Visual-Laser-Inertial SLAM Using a Compact 3D Scanner for Confined Space

  • Daqian Cheng
  • Haowen Shi
  • Albert Xu
  • Michael Schwerin
  • Michelle Crivella
  • Lu Li
  • Howie Choset

Three-dimensional reconstruction in confined spaces is important for the manufacturing of aircraft wings, inspection of narrow pipes, examination of turbine blades, etc. It is also challenging because confined spaces tend to lack a positioning infrastructure, and conventional sensors often cannot detect objects in close range. Therefore, such tasks require a sensor that is compact, operates in short-range, and able to localize itself. In this paper, we introduce a miniature and low-cost 3D scanning system including an active laser-stripe triangulation hardware, integrated inertial sensors, and a Simultaneous Localization and Mapping (SLAM) software tailored for the sensor. The proposed system is capable of reconstructing photo-realistic 3D point cloud in real-time in spite of its compact monocular configuration. To achieve this capability, we propose an approach to capture both color and geometry using alternating shutter-speed on a single camera. A novel SLAM method is proposed to accurately localize the sensor by fusing laser, camera, and inertial measurements. Evaluation of localization accuracy and comparison on reconstruction performance against a significantly larger commercial off-the-shelf sensor demonstrate the proposed system’s advantages in real-world applications.

Details

ICRA Conference 2019 Conference Paper

A Multi-Domain Feature Learning Method for Visual Place Recognition

  • Peng Yin 0001
  • Lingyun Xu
  • Xueqian Li
  • Chen Yin
  • Yingli Li
  • Rangaprasad Arun Srivatsan
  • Lu Li
  • Jianmin Ji

Visual Place Recognition (VPR) is an important component in both computer vision and robotics applications, thanks to its ability to determine whether a place has been visited and where specifically. A major challenge in VPR is to handle changes of environmental conditions including weather, season and illumination. Most VPR methods try to improve the place recognition performance by ignoring the environmental factors, leading to decreased accuracy decreases when environmental conditions change significantly, such as day versus night. To this end, we propose an end-to-end conditional visual place recognition method. Specifically, we introduce the multi-domain feature learning method (MDFL) to capture multiple attribute-descriptions for a given place, and then use a feature detaching module to separate the environmental condition-related features from those that are not. The only label required within this feature learning pipeline is the environmental condition. Evaluation of the proposed method is conducted on the multi-season NORDLAND dataset, and the multi-weather GTAV dataset. Experimental results show that our method improves the feature robustness against variant environmental conditions.

Details

ICRA Conference 2019 Conference Paper

MRS-VPR: a multi-resolution sampling based global visual place recognition method

  • Peng Yin 0001
  • Rangaprasad Arun Srivatsan
  • Yin Chen
  • Xueqian Li
  • Hongda Zhang
  • Lingyun Xu
  • Lu Li
  • Zhenzhong Jia

Place recognition and loop closure detection are challenging for long-term visual navigation tasks. SeqSLAM is considered to be one of the most successful approaches to achieve long-term localization under varying environmental conditions and changing viewpoints. SeqSLAM uses a brute-force sequential matching method, which is computationally intensive. In this work, we introduce a multi-resolution sampling-based global visual place recognition method (MRS-VPR), which can significantly improve the matching efficiency and accuracy in sequential matching. The novelty of this method lies in the coarse-to-fine searching pipeline and a particle filter-based global sampling scheme, that can balance the matching efficiency and accuracy in the long-term navigation task. Moreover, our model works much better than SeqSLAM when the testing sequence is over a much smaller time scale than the reference sequence. Our experiments demonstrate that MRSVPR is efficient in locating short temporary trajectories within long-term reference ones without compromising on the accuracy compared to SeqSLAM.

Details

IROS Conference 2018 Conference Paper

Stabilize an Unsupervised Feature Learning for LiDAR-based Place Recognition

  • Peng Yin 0001
  • Lingyun Xu
  • Zhe Liu 0022
  • Lu Li
  • Hadi Salman
  • Yuqing He
  • Weiliang Xu 0001
  • Hesheng Wang 0001

Place recognition is one of the major challenges for the LiDAR-based effective localization and mapping task. Traditional methods are usually relying on geometry matching to achieve place recognition, where a global geometry map need to be restored. In this paper, we accomplish the place recognition task based on an end-to-end feature learning framework with the LiDAR inputs. This method consists of two core modules, a dynamic octree mapping module that generates local 2D maps with the consideration of the robot's motion; and an unsupervised place feature learning module which is an improved adversarial feature learning network with additional assistance for the long-term place recognition requirement. More specially, in place feature learning, we present an additional Generative Adversarial Network with a designed Conditional Entropy Reduction module to stabilize the feature learning process in an unsupervised manner. We evaluate the proposed method on the Kitti dataset and North Campus Long-Term LiDAR dataset. Experimental results show that the proposed method outperforms state-of-the-art in place recognition tasks under long-term applications. What's more, the feature size and inference efficiency in the proposed method are applicable in real-time performance on practical robotic platforms.

Details

IROS Conference 2017 Conference Paper

Development of an inexpensive tri-axial force sensor for minimally invasive surgery

  • Lu Li
  • Bocheng Yu
  • Chen Yang
  • Prasad Vagdargi
  • Rangaprasad Arun Srivatsan
  • Howie Choset

This work presents the design and evaluation of a low-cost tri-axial force sensor, that has been developed to regain the sense of touch in minimally invasive surgeries (MIS). The force sensor uses an array of force sensitive resistors (FSR) with a mechanically pre-loaded structure to perform the force sensing. The sensor has a built-in signal conditioning circuitry to provide on-board power regulation, programmable signal amplification and analog to digital conversion. The sensor is inexpensive and highly sensitive to low-amplitude force, critical in surgical applications. We validate the efficacy of the sensor with two surgical applications - robotic palpation for stiffness mapping and obstacle avoidance for a highly articulated robotic probe (HARP). The results show that the sensor is capable of accurately detecting the stiff inclusions embedded in the tissues as well as detecting obstacles and helping HARP safely navigate around them.

Details
v2026.09.13