Multi-object tracking (MOT) in water surface scenes is crucial for the autonomous navigation of Unmanned Surface Vehicles (USVs). However, existing MOT datasets rarely focus on these scenes. Moreover, the few available water surface MOT datasets contain limited data shot onboard and concentrate narrowly on specific marine scenes, creating a significant gap from real-world USV navigation applications. To promote research on USV autonomous navigation, we introduce USVTrack, a fully onboard-shot MOT benchmark that covers diverse and complex water surface scenes, characterized by a high proportion of small objects and varied backgrounds. Then, we propose an innovative end-to-end method specifically designed for MOT in complex water surface scenes, termed as USVMOT. It improves tracking performance through four key contributions: 1) integrating mask information via knowledge distillation to boost feature discriminability; 2) deploying task-specific auxiliary pathways to alleviate the competition between detection and re-identification (ReID) in end-to-end MOT methods; 3) employing an adaptive high-quality mask generation strategy based on the Segment Anything Model (SAM) that obviates extensive manual annotation; and 4) introducing an object-aware association method that dynamically tailors the tracking strategy according to object size and motion speed. Extensive experiments on the USVTrack benchmark demonstrate that USVMOT outperforms existing methods. Our analysis reveals that MOT in complex water surface scenes remains challenging, highlighting the need for further advancements.

This paper studies Learning from Imperfect Human Feedback (LIHF), addressing the potential irrationality or imperfect perception when learning from comparative human feedback. Building on evidences that human's imperfection decays over time (i.e., humans learn to improve), we cast this problem as a concave-utility continuous-action dueling bandit but under a restricted form of corruption: i.e., the corruption scale is decaying over time as $t^{\rho-1}$ for some ``imperfection rate'' $\rho \in [0, 1]$. With $T$ as the total number of iterations, we establish a regret lower bound of $ \Omega(\max\{\sqrt{T}, T^{\rho}\})$ for LIHF, even when $\rho$ is known. For the same setting, we develop the Robustified Stochastic Mirror Descent for Imperfect Dueling (RoSMID) algorithm, which achieves nearly optimal regret $\tilde{\mathcal{O}}(\max\{\sqrt{T}, T^{\rho}\})$. Core to our analysis is a novel framework for analyzing gradient-based algorithms for dueling bandit under corruption, and we demonstrate its general applicability by showing how this framework can be easily applied to obtain corruption-robust guarantees for other popular gradient-based dueling bandit algorithms. Our theoretical results are validated by extensive experiments.

We investigate the robustness of multi-agent learning in strongly monotone games with bandit feedback. While previous research has developed learning algorithms that achieve last-iterate convergence to the unique Nash equilibrium (NE) at a polynomial rate, we demonstrate that all such algorithms are vulnerable to adversaries capable of poisoning even a single agent's utility observations. Specifically, we propose an attacking strategy such that for any given time horizon $T$, the adversary can mislead any multi-agent learning algorithm to converge to a point other than the unique NE with a corruption budget that grows sublinearly in $T$. To further understand the inherent robustness of these algorithms, we characterize the fundamental trade-off between convergence speed and the maximum tolerable total utility corruptions for two example algorithms, including the state-of-the-art one. Our theoretical and empirical results reveal an intrinsic efficiency-robustness trade-off: the faster an algorithm converges, the more vulnerable it becomes to utility poisoning attacks. To the best of our knowledge, this is the first work to identify and characterize such a trade-off in the context of multi-agent learning.

Online advertising platforms use automated auctions to connect advertisers with potential customers, requiring effective bidding strategies to maximize profits. Accurate ad impact estimation requires considering three key factors: delayed and long-term effects, cumulative ad impacts such as reinforcement or fatigue, and customer heterogeneity. However, these effects are often not jointly addressed in previous studies. To capture these factors, we model ad bidding as a Contextual Markov Decision Process (CMDP) with delayed Poisson rewards. For efficient estimation, we propose a two-stage maximum likelihood estimator combined with data-splitting strategies, ensuring controlled estimation error based on the first-stage estimator's (in)accuracy. Building on this, we design a reinforcement learning algorithm to derive efficient personalized bidding strategies. This approach achieves a near-optimal regret bound of $\tilde{\mathcal{O}}(dH^2\sqrt{T})$, where $d$ is the contextual dimension, $H$ is the number of rounds, and $T$ is the number of customers. Our theoretical findings are validated by simulation experiments.

Autonomous surface vehicles (ASVs) are widely employed in various marine industrial tasks. Regardless of the specific operation, one common process is docking, particularly the ASV's return to port. Therefore, the docking process is crucial to ASV operations in industrial applications. Docking space perception (DSP) provides essential environmental information, such as the docking space, driving space, and the shore. However, only a few studies have utilized marks for cooperative docking space perception, and no research has explored noncooperative perception. This article proposes a novel docking space bird's eye view (DSBEV) network to segment docking space on the bird's eye view (BEV) plane for the DSP task. DSBEV does not require any marks and can operate directly in port environments. DSBEV network includes an undistortion-splat module and an attention-water module, designed to enhance the DSP performance. A new watershore loss is introduced to generate the docking space utilizing the shape information. Due to the lack of the docking perception dataset, we construct a water surface platform to create a docking space dataset. The experimental results demonstrate that our DSBEV network achieved an average IoU of 76.8%, attaining state-of-the-art performance. Module analyses and ablation studies are conducted to validate the effectiveness of our designed modules.

Moving object segmentation (MOS) and Ego velocity estimation (EVE) are vital capabilities for mobile systems to achieve full autonomy. Several approaches have attempted to achieve MOSEVE using a LiDAR sensor. However, LiDAR sensors are typically expensive and susceptible to adverse weather conditions. Instead, millimeter-wave radar (MWR) has gained popularity in robotics and autonomous driving for real applications due to its cost-effectiveness and resilience to bad weather. Nonetheless, publicly available MOSEVE datasets and approaches using radar data are limited. Some existing methods adopt point convolutional networks from LiDAR-based approaches, ignoring the specific artifacts and the valuable radial velocity information of radar measurements, leading to suboptimal performance. In this paper, we propose a novel transformer network that effectively addresses the sparsity and noise issues and leverages the radial velocity measurements of radar points using our devised radar self- and cross-attention mechanisms. Based on that, our method achieves accurate EVE of the robot and performs MOS using only radar data simultaneously. To thoroughly evaluate the MOSEVE performance of our method, we annotated the radar points in the public View-of-Delft (VoD) dataset and additionally constructed a new radar dataset in various environments. The experimental results demonstrate the superiority of our approach over existing state-of-the-art methods. The code is available at https://github.com/ORCAUboat/RadarMOSEVE.

The millimeter-wave radar sensor maintains stable performance under adverse environmental conditions, making it a promising solution for all-weather perception tasks, such as outdoor mobile robotics. However, the radar point clouds are relatively sparse and contain massive ghost points, which greatly limits the development of mmWave radar technology. In this paper, we propose a novel point cloud super-resolution approach for 3D mmWave radar data, named Radar-diffusion. Our approach employs the diffusion model defined by mean-reverting stochastic differential equations (SDE). Using our proposed new objective function with supervision from corresponding LiDAR point clouds, our approach efficiently handles radar ghost points and enhances the sparse mmWave radar point clouds to dense LiDAR-like point clouds. We evaluate our approach on two different datasets, and the experimental results show that our method outperforms the state-of-the-art baseline methods in 3D radar super-resolution tasks. Furthermore, we demonstrate that our enhanced radar point cloud is capable of downstream radar point-based registration tasks.

Monocular visual odometry (VO) has extensive applications in mobile robots and computer vision. However, current applications of monocular VO systems in complex environments still have limitations. Accurate, robust, and easy-to-use VO is still an unsolved problem to some extent. In recent years, the single-chip millimeter-wave (mmWave) radar has been increasingly used in various types of mobile robots due to its advantages of small size, low cost, and robustness in harsh weather conditions. In this paper, we apply the mmWave radar to a VO system and propose a multi-stage visual-radar fusion odometry framework, MS-VRO. The framework is based on a typical monocular VO system. By merging mmWave radar data in different stages, the proposed odometry improves the accuracy, robustness, and generalization ability of VO. The framework contains a new visual-radar initialization method, a visual-radar joint optimization method, and a radar-aided visual feature selection and processing method that can remove dynamic object features and bad map points. Through these, the proposed method solves the problems of monocular VO, including scale ambiguity, scale drift, and performance degradation in dynamic environments. We build a dataset that can be used for research on visual-radar fusion odometry and test the proposed method on the new dataset and other public datasets. The result shows that the proposed odometry achieves significantly better performance than VO methods and is more accurate and robust compared to some typical visual-inertial odometry methods.

Person reidentification (ReID) systems play a key role in intelligent visual surveillance systems and have widespread applications, for example, in public security. Usually, person ReID systems can identify a person with cameras. In this article, we focus on the relatively unexplored area of using low-cost automotive radar for the person ReID problem. Unlike the radar-based person identification, person ReID has some characteristics, such as the uncooperative scenes and the long-term robustness. Therefore, we design a new deep learning network to extract spatiotemporal information from 4-D radar point clouds. We also build a data set of radar point clouds collected from the real-world person ReID scenarios. The evaluation result shows that our method achieves 91% CMC-1 accuracy on the ReID task. Besides, for the person identification task, our method also achieves accuracies of 98% and 91% for 15 and 40 individuals, respectively. In addition, we discuss the potential of using radar for person ReID problems and intuitively explain the new method’s performance. Finally, we analyze the robustness and the influence of different parameters on the method and the contributions of different modules to the network model. The results of our experiment indicate that radar-based ReID not only preserves privacy but also outperforms camera-based ReID in some cases, such as in low-light environments or with substantial clothing changes.

Millimeter-wave (mmWave) radar has been widely used in autonomous driving due to its good performance under harsh weather conditions. In recent years, with the development of mmWave radar hardware performance, radar point clouds, as an important data format of mmWave radar, have been widely used in high-level perception tasks of mobile robots and autonomous driving. However, at present, compared to LiDAR point clouds, in common application scenes of mobile robots, mmWave radar point clouds have shortcomings such as sparsity and containing many “ghost” targets. Therefore, in this article, we analyze the reasons that cause these problems and propose a new method for point cloud generation as well as a new evaluation metric. After building a new dataset and carrying out experiments in real-world scenes, our method shows better performance on the quality of radar point clouds compared to other methods. In addition, by evaluating the performance of applying the high-quality radar point clouds to object detection tasks as well as localization and mapping tasks, the result shows that radar point clouds generated using our method can significantly improve the environment perception ability of mobile robots.

The millimeter-wave radar, as an important sensor, is widely used in autonomous driving. In recent years, to meet the requirement of high level autonomous driving applications, attentions have been paid to generate high-quality radar point clouds. However, in the complex roadway environment, the weaknesses of classical radar detectors are exposed, such as too much clutter points and sparse valid point clouds. Therefore, in this paper, we propose a new automotive radar detector based on deep learning using the spatial distribution feature of the real targets, in order to improve the performance of automotive radar detector in the real-world driving scene. Besides, aiming at the lack of radar data labels, we propose an autonomous labeling method by using synchronized Lidar data. Finally, we evaluate the detector on data collected in the real-world roadway scene and the result shows that the proposed radar detector out-performs the classical radar detectors in suppressing the clutter and generating denser point clouds.

In recent years, unmanned surface vehicles (USVs) have been experiencing growth in various applications. With the expansion of USVs’ application scenes from the typical marine areas to inland waters, new challenges arise for the object detection task, which is an essential part of the perception system of USVs. In our work, we focus on a relatively unexplored task for USVs in inland waters: small object detection on water surfaces, which is of vital importance for safe autonomous navigation and USVs’ certain missions such as floating waste cleaning. Considering the limitations of vision-based object detection, we propose a novel radar-vision fusion based method for robust small object detection on water surfaces. By using a novel representation format of millimeter wave radar point clouds and applying a deep-level multi-scale fusion of RGB images and radar data, the proposed method can efficiently utilize the characteristics of radar data and improve the accuracy and robustness for small object detection on water surfaces. We test the method on the real-world floating bottle dataset that we collected and released. The result shows that, our method improves the average detection accuracy significantly compared to the vision-based methods and achieves state-of-the-art performance. Besides, the proposed method performs robustly when single sensor degrades.

Marine debris is severely threatening the marine lives and causing sustained pollution to the whole ecosystem. To prevent the wastes from getting into the ocean, it is helpful to clean up the floating wastes in inland waters using the autonomous cleaning devices like unmanned surface vehicles. The cleaning efficiency relies on a high-accurate and robust object detection system. However, the small size of the target, the strong light reflection over water surface, and the reflection of other objects on bank-side all bring challenges to the vision-based object detection system. To promote the practical application for autonomous floating wastes cleaning, we present FloW†, the first dataset for floating waste detection in inland water areas. The dataset consists of an image sub-dataset FloW-Img and a multimodal sub-dataset FloW-RI which contains synchronized millimeter wave radar data and images. Accurate annotations for images and radar data are provided, supporting floating waste detection strategies based on image, radar data, and the fusion of two sensors. We perform several baseline experiments on our dataset, including vision-based and radar-based detection methods. The results show that, the detection accuracy is relatively low and floating waste detection still remains a challenging task.

BACKGROUND: Identifying key "driver" mutations which are responsible for tumorigenesis is critical in the development of new oncology drugs. Due to multiple pharmacological successes in treating cancers that are caused by such driver mutations, a large body of methods have been developed to differentiate these mutations from the benign "passenger" mutations which occur in the tumor but do not further progress the disease. Under the hypothesis that driver mutations tend to cluster in key regions of the protein, the development of algorithms that identify these clusters has become a critical area of research. RESULTS: We have developed a novel methodology, QuartPAC (Quaternary Protein Amino acid Clustering), that identifies non-random mutational clustering while utilizing the protein quaternary structure in 3D space. By integrating the spatial information in the Protein Data Bank (PDB) and the mutational data in the Catalogue of Somatic Mutations in Cancer (COSMIC), QuartPAC is able to identify clusters which are otherwise missed in a variety of proteins. The R package is available on Bioconductor at: http://bioconductor.jp/packages/3.1/bioc/html/QuartPAC.html . CONCLUSION: QuartPAC provides a unique tool to identify mutational clustering while accounting for the complete folded protein quaternary structure.

BACKGROUND: Current research suggests that a small set of "driver" mutations are responsible for tumorigenesis while a larger body of "passenger" mutations occur in the tumor but do not progress the disease. Due to recent pharmacological successes in treating cancers caused by driver mutations, a variety of methodologies that attempt to identify such mutations have been developed. Based on the hypothesis that driver mutations tend to cluster in key regions of the protein, the development of cluster identification algorithms has become critical. RESULTS: We have developed a novel methodology, SpacePAC (Spatial Protein Amino acid Clustering), that identifies mutational clustering by considering the protein tertiary structure directly in 3D space. By combining the mutational data in the Catalogue of Somatic Mutations in Cancer (COSMIC) and the spatial information in the Protein Data Bank (PDB), SpacePAC is able to identify novel mutation clusters in many proteins such as FGFR3 and CHRM2. In addition, SpacePAC is better able to localize the most significant mutational hotspots as demonstrated in the cases of BRAF and ALK. The R package is available on Bioconductor at: http://www.bioconductor.org/packages/release/bioc/html/SpacePAC.html. CONCLUSION: SpacePAC adds a valuable tool to the identification of mutational clusters while considering protein tertiary structure.

BACKGROUND: It is well known that the development of cancer is caused by the accumulation of somatic mutations within the genome. For oncogenes specifically, current research suggests that there is a small set of "driver" mutations that are primarily responsible for tumorigenesis. Further, due to recent pharmacological successes in treating these driver mutations and their resulting tumors, a variety of approaches have been developed to identify potential driver mutations using methods such as machine learning and mutational clustering. We propose a novel methodology that increases our power to identify mutational clusters by taking into account protein tertiary structure via a graph theoretical approach. RESULTS: We have designed and implemented GraphPAC (Graph Protein Amino acid Clustering) to identify mutational clustering while considering protein spatial structure. Using GraphPAC, we are able to detect novel clusters in proteins that are known to exhibit mutation clustering as well as identify clusters in proteins without evidence of prior clustering based on current methods. Specifically, by utilizing the spatial information available in the Protein Data Bank (PDB) along with the mutational data in the Catalogue of Somatic Mutations in Cancer (COSMIC), GraphPAC identifies new mutational clusters in well known oncogenes such as EGFR and KRAS. Further, by utilizing graph theory to account for the tertiary structure, GraphPAC discovers clusters in DPP4, NRP1 and other proteins not identified by existing methods. The R package is available at: http://bioconductor.org/packages/release/bioc/html/GraphPAC.html. CONCLUSION: GraphPAC provides an alternative to iPAC and an extension to current methodology when identifying potential activating driver mutations by utilizing a graph theoretic approach when considering protein tertiary structure.

BACKGROUND: Human cancer is caused by the accumulation of somatic mutations in tumor suppressors and oncogenes within the genome. In the case of oncogenes, recent theory suggests that there are only a few key "driver" mutations responsible for tumorigenesis. As there have been significant pharmacological successes in developing drugs that treat cancers that carry these driver mutations, several methods that rely on mutational clustering have been developed to identify them. However, these methods consider proteins as a single strand without taking their spatial structures into account. We propose an extension to current methodology that incorporates protein tertiary structure in order to increase our power when identifying mutation clustering. RESULTS: We have developed iPAC (identification of Protein Amino acid Clustering), an algorithm that identifies non-random somatic mutations in proteins while taking into account the three dimensional protein structure. By using the tertiary information, we are able to detect both novel clusters in proteins that are known to exhibit mutation clustering as well as identify clusters in proteins without evidence of clustering based on existing methods. For example, by combining the data in the Protein Data Bank (PDB) and the Catalogue of Somatic Mutations in Cancer, our algorithm identifies new mutational clusters in well known cancer proteins such as KRAS and PI3KC α. Further, by utilizing the tertiary structure, our algorithm also identifies clusters in EGFR, EIF2AK2, and other proteins that are not identified by current methodology. The R package is available at: http://www.bioconductor.org/packages/2.12/bioc/html/iPAC.html. CONCLUSION: Our algorithm extends the current methodology to identify oncogenic activating driver mutations by utilizing tertiary protein structure when identifying nonrandom somatic residue mutation clusters.