Cloud robotics enables robots to offload computationally intensive tasks to cloud servers for performance, cost, and ease of management. However, the network and cloud computing infrastructure are not designed for reliable timing guarantees, due to fluctuating Quality-of-Service (QoS). In this work, we formulate an impossibility triangle theorem for: Latency reliability, Singleton server, and Commodity hardware. The LSC theorem suggests that providing replicated servers with uncorrelated failures can exponentially reduce the probability of missing a deadline. We present FogROS2-Probabilistic Latency Reliability (PLR) that uses multiple independent network interfaces to send requests to replicated cloud servers and uses the first response back. We design routing mechanisms to discover, connect, and route through non-default network interfaces on robots. FogROS2-PLR optimizes the selection of interfaces to servers to minimize the probability of missing a deadline. We conduct a cloud-connected driving experiment with two 5 G service providers, demonstrating FogROS2-PLR effectively provides smooth service quality even if one of the service providers experiences low coverage and base station handover. We use 99 Percentile (P99) latency to evaluate anomalous long-tail latency behavior. In one experiment, FogROS2-PLR improves P99 latency by up to 3.7 x compared to using one service provider. We deploy FogROS2-PLR on a physical Stretch 3 robot performing an indoor human-tracking task. Even in a fully covered$\text{Wi}-\text{Fi}$and 5 G environment, FogROS2-PLR improves the responsiveness of the robot reducing mean latency by 36% and P99 latency by 33%. Code and supplementary can be found on website11https://github.com/data-capsule/rt-fogros2.

Generative AI systems have shown impressive capabilities in creating text, code, and images. Inspired by the importance of research in industrial Design for Assembly, we introduce a novel problem: Generative Design-for-RobotAssembly (GDfRA). The task is to generate an assembly based on a natural language prompt (e.g., “giraffe”) and an image of available physical components, such as 3D-printed blocks. The output is an assembly, a spatial arrangement of these components, accompanied by instructions for a robot to build it. The output geometry must 1) resemble the requested object and 2) be reliably assembled by a 6 DoF robot arm with a suction gripper. We then present Blox-Net, a GDfRA system that combines generative vision language models with well-established methods in computer vision, simulation, perturbation analysis, motion planning, and physical robot experimentation to solve a class of GDfRA problems without human supervision. Blox-Net achieved a Top-1 accuracy of$\mathbf{6 3. 5 \%}$in the semantic accuracy of its designed assemblies. Six designs, after Blox-Net's automated pertubation redesign, were reliably assembled by a robot, achieving near-perfect success across$\mathbf{1 0}$consecutive assembly iterations with human intervention only during reset prior to assembly. The entire pipeline from the textual word to reliable physical assembly is performed without human intervention. Project Page: https://bloxnet.org/

3D Gaussian Splats (3DGSs) are 3D object models derived from multi-view images. Such “digital twins” are useful for simulations, virtual reality, E-commerce, robot policy fine-tuning, and part inspection. 3D object scanning usually requires multi-camera arrays, precise laser scanners, or robot wrist-mounted cameras, which have restricted workspaces. We propose Omni-Scan, a pipeline for producing high-quality 3D Gaussian Splat models using a bi-manual robot that grasps an object with one gripper and rotates the object with respect to one stationary camera. The object is then re-grasped by a second gripper to expose surfaces that were occluded by the first gripper. We present the Omni-Scan robot pipeline using DepthAnything, Segment Anything, as well as RAFT optical flow models to identify and isolate objects held by a robot gripper while removing the gripper and the background. We then modify the 3DGS training pipeline to support concatenated datasets with gripper occlusion, producing an omni-directional (360°) model of the object. We apply Omni-Scan to part defect inspection, finding that it can identify visual or geometric defects in 12 different industrial and household objects with an average accuracy of 83.3%. More details and interactive videos of Omni-Scan 3DGS models can be found at https://berkeleyautomation.github.io/omni-scan/.

In logistics, the ability to quickly compute and execute pick-and-place motions from bins is critical to increasing productivity. We present Bin-Optimized Motion Planning (BOMP), a motion planning framework that plans arm motions for a six-axis industrial robot with a long-nosed suction tool to remove boxes from deep bins. BOMP considers robot arm kinematics, actuation limits, the dimensions of a grasped box, and a varying height map of a bin environment to rapidly generate time-optimized, jerk-limited, and collision-free trajectories. The optimization is warm-started using a deep neural network trained offline in simulation with 25,000 scenes and corresponding trajectories. Experiments with 96 simulated and 15 physical environments suggest that BOMP generates collision-free trajectories that are up to 58% faster than baseline sampling-based planners and up to 36% faster than an industry-standard Up-Over-Down algorithm, which has an extremely low 15% success rate in this context. BOMP also generates jerk-limited trajectories while baselines do not. Website: https://sites.google.com/berkeley.edu/bomp.

Integrating cognitive radio into Internet of Things (IoT) is conducive to reducing spectrum scarcity for large-scale IoT deployment, where a core technology is the design of spectrum access algorithms for effective assignment of spectrum holes. However, due to the partially observable channels and increased number of users in the cognitive radio Internet of Things (CRIoT) network, the secondary users have difficulty avoiding interferences and accessing the spectrum quickly. This study presents a distributed dynamic spectrum access (DSA) algorithm that employs a priority experience replay deep echo state Q-network (PER-DESQN) for CRIoT networks with multiple users and channels. To accelerate the Q-network convergence, we use an echo state network based on the underlying temporal correlation to estimate Q-values. Then, to resolve the Q-value overestimation and improve prediction accuracy, the estimated Q-value and decision action process are trained using a double deep Q-network (DDQN). Moreover, a priority experience replay mechanism that uses the Sum-Tree combined with importance sampling weights is proposed to optimize the DDQN to address the instability of the Q-value resulting from random sampling. As the simulation results demonstrate, the proposed algorithm can make fast and accurate DSA decisions and boost the network channel capacity significantly.

We consider a decluttering problem where multiple rigid convex polygonal objects rest in randomly placed positions and orientations on a planar surface and must be efficiently transported to a packing box using both single and multi-object grasps. Prior work considered frictionless multi-object grasping. In this paper, we introduce friction to increase the number of potential grasps for a given group of objects, and thus increase picks per hour. We train a neural network using real examples to plan robust multi-object grasps. In physical experiments, we find a 13.7% increase in success rate, a 1.6x increase in picks per hour, and a 6.3x decrease in grasp planning time compared to prior work on multi-object grasping. Compared to single-object grasping, we find a 3.1x increase in picks per hour.

Indoor positioning technology, as a crucial foundation of location-based services, is experiencing a growing need for high precision driven by the Internet of Things (IoT). However, traditional positioning algorithms suffer from low sample utilization and susceptibility to noise. Moreover, the presence of indoor obstacles significantly affects positioning accuracy and leads to the issue of wall-penetrating positioning. To address these problems, this paper proposes a hybrid time-of-arrival/angle-of-arrival (TOA/AOA) indoor positioning algorithm based on sparse reconstruction and particle filtering-based map matching. Specifically, sparse reconstruction is employed to improve the utilization of samples, and iterative updating of the position estimation is performed during the multi-sample joint estimation process to enhance accuracy. Furthermore, to tackle the problem of wall-penetrating positioning, a particle filtering-based map matching algorithm is proposed to detect and eliminate the wall-penetrating particles using the map information matrix, which optimizes the positioning results obtained from sparse reconstruction. Simulation results demonstrate the effectiveness of the proposed algorithm in satisfying the demand for high-precision indoor positioning.

The millimeter-wave (mmwave) massive multiple input multiple output (MIMO) system in beamspace can greatly reduce the number of radio frequency (RF) links through beam selection. However, the limited number of RF links make it very difficult to obtain high-dimensional beam space channel state information. Aiming at the problems of high complexity and low accuracy of beamspace channel estimation in mmwave massive MIMO system, a block support detection (BSD) algorithm based on block sparsity is proposed. Specifically, an equivalent vector with block sparsity is obtained by recombining the path components of the channel vector, then the non-zero elements of each path component are treated together to obtain the block sparsity support set. Thus, the high-dimensional beamspace channel can be estimated under the condition of low pilot overhead. Theoretical analysis and simulation results show that the proposed BSD algorithm is superior to some existing schemes in terms of estimation accuracy, pilot overhead and complexity.

Blind modulation classification has emerged as a promising technology in many military and civilian applications, such as cognitive radios, satellite systems, etc. However, it is very challenging to support this blind mechanism within non-Gaussian noise environments, which recently have been identified in a variety of electromagnetic communication networks. Also, start-of-the-art classification methods are mainly based on neural networks or deep learning, which inevitably induces heavy computation loads and thus cannot proactively learn from wireless data in real time. To address the challenges, this paper introduces a series of low- computation radio frequency analytics, including generalized cyclic spectrum (GCS), principal component analysis (PCA), and support vector machine (SVM), which enables the blind modulation classification under non-Gaussian noise. First, based on raw sensory signals and the designed bounded nonlinear function, GCS is extracted as the radio frequency feature to facilitate discrimination of modulation schemes. This GCS can also effectively suppress the burstiness impact of non-Gaussian noise. Then, PCA method is adopted to optimally reduce the dimensionality of GCS features, and a simple and efficient SVM classifier is employed to identify the exact modulation of received signals. Both Monte Carlo simulations and real- data experiments confirm that the proposed design outperforms existing solutions with higher classification accuracy and robustness, i.e., at least 13\% improvement of recognition accuracy in very low (-2 dB) generalized signal-to-noise ratio scenario.

During the last few years, automatic modulation classification (AMC) has attracted widespread attention in both civilian and military applications. Conventional AMC schemes are primarily developed under Gaussian noise assumptions. However, recent empirical studies show that non-Gaussian noise has emerged in a variety of wireless networked systems. The bursty nature of non-Gaussian noise fundamentally challenges the applicability of the conventional AMC schemes. In order to improve the classification performance under non-Gaussian noise, in this paper, a novel modulation classification method is proposed by using cyclic correntropy spectrum (CCES) and deep residual neural network (ResNet). First, CCES is introduced to effectively suppress non-Gaussian noise through the designated Gaussian kernel. CCES also provides significantly different CCES graphs with respect to different modulation schemes, enabling AMC to directly operate with the graphs without further feature extraction. Next, based on the CCES graphs, an end-to-end deep ResNet-based AMC is developed to recognize the correct modulation by iteratively evaluating the residual information in a cascade of multiple learning layers. Experimental results confirm that the proposed algorithm outperforms existing designs with much higher classification accuracy, i.e., 3 dB less in the required generalized signal to noise ratio for 100% accuracy, in non-Gaussian noise environments.

Non-Gaussianity of noises and non-stationarity of signals have been the two crucial considerations in the fields of signal processing and communications. Correspondingly, many denoising and cyclostationary methods have been published to deal with the relevant problems, respectively. Recently, a novel method named cyclic correntropy or cyclostationary correntropy was proposed to deal with the two problems simultaneously. Thanks to the symmetry and sparsity of cyclic correntropy spectrum, the compressed spectrum obtained by compressive sensing can be used to complete the task in certain situations. In this letter, a novel method to estimate the cyclic frequency by compressed cyclic correntropy spectrum is proposed to reduce computational complexity and storage cost. To reveal the proposed method's effectiveness and robustness to impulsive noise, a number of numerical experiments are carried out to compare with existing cyclostationary methods. As cyclostationary signal processing and compressive sensing are the two great theories, through in-depth study, they can have more collaborative work opportunities and meanings.

In this paper, an adaptive filtering algorithm, termed the extended kernel recursive maximum correntropy (EX-KRMC) algorithm is proposed as a novel approach of traditional recursion based adaptive filtering algorithms. Maximum correntropy criterion is employed to better the robustness to non-Gaussian noise and kernel methods are used to enable the capacity for nonlinear systems. It is verified by simulation experiments that EX-KRMC outperforms existing adaptive filtering algorithms when dealing with non-Gaussian noise for nonlinear time-variant systems.

In this study, the authors propose a restoration algorithm for blurred and noisy continuous image sequences. The proposed approach treats an image sequence as a space‐time volume and employs a spatio‐temporal mean curvature regularisation which is a novel regularisation proposed in the study to enhance the smoothness of the solution. An augmented Lagrangian method with splitting techniques is used to handle the problem, iteratively finding solutions to the subproblems. Experiments show that the proposed approach can produce higher quality results and more natural images comparing with other space‐time volume based methods on image sequence denoising and deblurring problems.

In this paper, we propose a novel approach for performing high-quality edge-preserving image filtering. Based on a local linear model and using the principle of Stein's unbiased risk estimate as an estimator for the mean squared error from the noisy image only, we derive a simple explicit image filter which can filter out noise while preserving edges and fine-scale details. Moreover, this filter has a fast and exact linear-time algorithm whose computational complexity is independent of the filtering kernel size; thus, it can be applied to real time image processing tasks. The experimental results demonstrate the effectiveness of the new filter for various computer vision applications, including noise reduction, detail smoothing and enhancement, high dynamic range compression, and flash/no-flash denoising.

In this study, the authors focus on the tracking performance and the robustness of 12 sliding-windowed Hurst estimators for multifractional processes with linear trend local Hölder exponent, noisy multifractional processes and multifractional processes with infinite second-order statistics. Four types of multifractional processes are synthesised to test the tracking performance and robustness of these 12 sliding-windowed Hurst estimators. They are (i) noise-free multifractional process; (ii) multifractional process corrupted by 30-dB signal-to-noise ratio (SNR) white Gaussian noise; (iii) multifractional process corrupted by 30-dB SNR impulse noise; and (iv) multifractional stable process, which has no finite second-order statistics. Furthermore, the standard error of different sliding-windowed Hurst estimators are calculated in order to quantify the accuracy and robustness. This study provides a guideline and principle in the selection of Hurst estimators for noise-free multifractional process, noise-corrupted multifractional process and multifractional process with infinite second-order statistics. The results of this analysis show that the sliding-windowed Kettani and Gubner's method provides the best-tracking performance for multifractional processes with linear trend local Hölder exponent and good robustness to noise.

The minimum error entropy (MEE) criterion in information theoretic learning is an efficient way to deal with non-Gaussian signal processing. And the minimum dispersion (MD) criterion has been widely applied in stable signal processing. In this letter, we show that there exists an equivalence between the MD criterion and the MEE criterion where symmetric$\alpha$-stable$(S\alpha S)$random variables are considered as the errors of the adaptive signal processing. As an application, we propose an algorithm with the MEE criterion for the time delay estimation (TDE) problem which was solved by the MD criterion.