Search

This article presents a comprehensive survey of the integration of machine learning techniques into robotic grasping, with a special emphasis on the challenges and advancements for space applications. The incorporation of artificial intelligence, particularly through deep learning, reinforcement learning, transfer learning, convolutional neural networks and recurrent neural networks, has significantly revolutionized robotic grasping. These advancements facilitate autonomous, efficient, and sophisticated manipulation in the challenging environment of outer space, transitioning from traditional mechanical grippers to sophisticated systems powered by advanced algorithms. This transition highlights the critical integration of sensory perception, grasp planning, and execution mechanisms, enhancing robots' capabilities to perceive, interact with, and manipulate objects with unprecedented precision and adaptability. The article meticulously outlines significant advancements achieved through the deployment of convolutional neural networks for visual information processing, RNNs for sequential decision-making, RL for autonomous strategy refinement, and transfer learning for leveraging pre-learned knowledge in novel tasks. These technologies address the unique challenges of space environments, such as varied textures, occlusions, microgravity conditions, and the sim-to-real gap, by enhancing sample efficiency, improving sim-to-real transfer capabilities, and integrating multimodal data for better object localization and pose estimation. Furthermore, the review explores the specific challenges faced in space robotic grasping, including handling varied textures and occlusions, adapting to unpredictable conditions, achieving real-time processing, and ensuring safety and reliability. It proposes future research directions focused on overcoming these hurdles, such as enhanced generalization through multimodal learning, robust sim-to-real transfer techniques, and the development of collaborative robotics and swarm intelligence. Critical to the development of ML models for robotic grasping are the roles of specialized datasets and simulation environments. Datasets like the Cornell Grasping Dataset and the Yale-CMU-Berkeley Object, along with simulation platforms such as Gazebo and PyBullet, provide essential resources for training, testing, and refining ML models. These tools enable researchers to simulate complex robotic systems and interactions within realistic environments, fostering rapid iterations on design and control strategies. In summary, this article offers in-depth insights into the progress, current challenges, and future prospects of machine learning techniques in robotic grasping for space exploration. It showcases significant strides made in the field and charts a path forward, emphasizing the need for innovative solutions to navigate the complexities of robotic manipulation in outer space. Through the strategic integration of advanced ML techniques, the development of adaptable and efficient robotic systems for space applications continues to advance, promising to unlock new possibilities in space exploration and beyond. • Surveys machine learning's application in space robotic grasping, noting key advances. • Highlights CNNs and RNNs' roles in robotic perception and action. • Discusses deep learning's evolution in autonomous space robotics. • Outlines future directions, emphasizing sim-to-real and multimodal learning. [ABSTRACT FROM AUTHOR]

In this paper, a unified neural control scheme is presented for the output-constrained trajectory tracking of space manipulator under unknown parameters and external perturbations. By utilizing the backstepping control technique as the main framework, the proposed controller is developed with the help of neural network (NN) and tan-type barrier Lyapunov function (BLF). The NN is introduced to identify the unknown part in the dynamic model of the space manipulator. Benefiting from the neural identification, the proposed controller is model-free and insensitive to external perturbations. Moreover, the BLF is adopted to guarantee the position tracking errors never exceed the predefined output constraints. Different from log-type BLF, the tan-type BLF is employed for the control design, which makes the proposed controller universal for the cases with and without considering the output constraints. The semiglobal uniform ultimate boundedness of the resulting closed-loop system is strictly obtained through stability argument. All error variables in the closed-loop system can eventually stabilize to the small residual sets about zero under the proposed controller. Lastly, simulations and comparisons are given to demonstrate the effectiveness and excellent tracking performance of the proposed control scheme. [ABSTRACT FROM AUTHOR]

Abstract In this paper, we investigate some unrevealed features of a newly introduced megastable chaotic oscillator. This oscillator has a rich dynamical behavior including limit cycle, torus and strange attractor. Also it has coexisting self excited and hidden attractors. Such multi-stable oscillator with coexisting self excited and hidden attractors is very rare in the literature. We have studied the oscillator dynamics using an analog circuit. Also a novel fuzzy-based robust and adaptive control method is designed to control this oscillator. Highlights • In this paper, we investigate some unrevealed features of a newly introduced megastable chaotic oscillator. • This oscillator has a rich dynamical behavior including limit cycle, torus and strange attractor. • Also it has coexisting self excited and hidden attractors. • We have studied the oscillator dynamics using an analog circuit. • In addition to, a novel fuzzy-based robust and adaptive control method is designed to control this oscillator. [ABSTRACT FROM AUTHOR]

• A new neural network-based terminal sliding mode technique is proposed to stabilize and synchronize fractional-order chaotic ecological systems in finite time. • The Chebyshev neural network is implemented to estimate unknown functions of the system. • Through the proposed Chebyshev neural network observer, the effects of external disturbances are fully taken into account. • The finite-time convergence of the closed-loop system, which is a new concept for ecological systems, is proven In the present study, a new neural network-based terminal sliding mode technique is proposed to stabilize and synchronize fractional-order chaotic ecological systems in finite-time. The Chebyshev neural network is implemented to estimate unknown functions of the system. Moreover, through the proposed Chebyshev neural network observer, the effects of external disturbances are fully taken into account. The weights of the Chebyshev neural network observer are adjusted based on adaptive laws. The finite-time convergence of the closed-loop system, which is a new concept for ecological systems, is proven. Then, the dependency of the system on the value of the fractional time derivatives is investigated. Lastly, the proposed control scheme is applied to the fractional-order ecological system. Through numerical simulations, the performance of the developed technique for synchronization and stabilization are assessed and compared with a conventional method. The numerical simulations strongly corroborate the effective performance of the proposed control technique in terms of accuracy, robustness, and convergence time for the unknown nonlinear system in the presence of external disturbances. [ABSTRACT FROM AUTHOR]

Complex industrial processes (CIPs) tend to run in multiple operating modes due to a variety of factors, such as the changes in load or raw material quality, random interferences from external environments, and the changing of set-points due to diverse personalized demands on product qualities. Successive stable operation modes will inevitably go through a transition stage that is extremely similar to abnormal conditions (faults). Traditional fault detection methods struggle to distinguish transition processes from faults, usually leading to high false alarm rates and resulting in frequent system fluctuations with low product quality or high production consumption. Therefore, this article proposes a well-performing CIP fault monitoring scheme with intelligent identification of transition processes. Firstly, a neighbor inconsistent pairs-based incremental attribute (or process variable) reduction approach is proposed. It extracts key process variables, which can effectively reflect the intrinsic dynamical characteristics of CIPs, from process data with high dimension, complexity, inconsistency and redundancy. Successively, an adaptive optimal sliding-window-based transition process identification approach is proposed, and the corresponding sub-phase-based fault monitoring criteria are established to realize the online fault identification of CIPs. Extensive confirmatory and comparative experiments on a numerical simulation system and the benchmark Tennessee Eastman process show that the proposed method can accurately identify the transition process, which effectively improves the fault monitoring performance of CIPs. • An incremental NIP-based attribute reduction method is proposed to fully express the dynamic characteristics of CIPs. • An adaptive sliding window division method is proposed to dynamically partition the working stage of process conditions. • A simple, yet effective rule-based identification procedure is established to distinguish the steady and transition states. [ABSTRACT FROM AUTHOR]

The tailings grade is a crucial indicator for controlling and optimizing mineral flotation processes. However, detecting/predicting tailings grade online faces challenges due to intricate physicochemical reactions in three-phase slurries with the strongly-coupled characteristics of the technical process. This paper introduces a novel multi-source features-fused two-stage tailings grade prediction model called MsFfTsGP by integrating multiple-source froth image sequences (FISs) with available traditional technical variables (TTVs, e.g., the reagent dosage and measured grades) from multiple flotation cells, mimicking the on-site two-stage prediction mode of experienced operators. Namely, an initial evaluation is made based on the source-ore characteristics using the FISs of the first rougher, while the second stage provides a sophisticated prediction of the tailing grade by considering both the source-ore characteristics and the flotation behavior of the whole circuit process. Specifically, FISs captured from the first rougher are fed into a multi-stream 3D convolutional network (Ms3DCN) for making an initial prediction. Then, the initial prediction result is combined with the FIS information from the last scavenger (also processed by the Ms3DCN) and available TTVs based on the attention mechanism to make a sophisticated prediction. Extensive experiments on a real industrial lead-zinc flotation process demonstrated a significant improvement of the proposed model in prediction accuracies. Integrating multisource FISs with TTVs from multiple coupled flotation cells shows great potential for achieving accurate tailings grade prediction, making MsFfTsGP a promising tool for enhancing the control and optimization of flotation processes. [ABSTRACT FROM AUTHOR]

In this article, a distributed adaptive fault-tolerant control scheme is presented for the attitude consensus of multiple flexible spacecraft formulated by partial differential equations (PDEs). Each spacecraft is supposed to be actuated by a control torque on the rigid hub and a control force on the tip payload. First, a finite-time distributed observer is constructed to estimate the virtual leader's state for each follower spacecraft under local communication. Then, an individual controller is exploited for each follower spacecraft, which consists of an adaptive boundary control torque law to track the recovered attitude and velocity and an adaptive boundary control force law to eliminate the elastic deformation. Particularly, the parametric adaptation laws are incorporated to compensate the actuator faults and perturbations. The attitude synchronization errors, velocity synchronization errors, and elastic deformations can asymptotically stabilize to zero under the introduced controller. This work owns the following two major novelties. One is introducing a unified distributed control framework for the PDE-based attitude consensus of multiple flexible spacecraft. Another is developing a novel parametric adaptation strategy to accommodate the actuator faults. Lastly, comparative simulations verify and highlight the main results. [ABSTRACT FROM AUTHOR]

This article presents a prescribed performance adaptive control scheme for the attitude tracking and vibration reduction of flexible spacecraft modeled by partial differential equations (PDEs). First, the original constrained problem is converted to an equivalent unconstrained one by adopting a prescribed error transformation. Then, a boundary control torque law and a boundary control force law are developed to track the desired attitude and reduce the vibration, respectively. Moreover, the parametric adaptation technique is integrated with the boundary control laws to estimate the disturbances. The closed-loop system is evaluated to be asymptotically stable through the Lyapunov stability theorem and LaSalle's invariance principle. The developed controller has two distinctive features as follows. One is the controller can guarantee the attitude tracking error and tip deformation always satisfy the prescribed performance constraints. Another is the controller can acquire the strong robustness owing to the feedforward disturbance compensation. Lastly, the derived results are demonstrated by simulated studies. [ABSTRACT FROM AUTHOR]

The quality and ratio of clinker, the fineness of cement are key factors affecting the strength of cement. In order to realize the target tracking control of cement strength, a self-learning fuzzy predictive control algorithm is proposed to calculate the adjustment variables of cement grinding process Considering the serious hysteresis of cement strength detection, the 3-day cement strength needs to be predicted, statistical models cannot actually guide the production for the loss of working conditions. In this paper, a 3-day strength prediction model is established based on step response which is trained by difference data. The cement grinding process is a complex industrial process which is composed of multiple sub-processes and the variables are coupled each other, according to the characteristic, a multi-variable distributed fuzzy control algorithm (MVDFC) is proposed, and the fuzzy rules are mined by Apriori algorithm. The verification results from the actual field data demonstrate the effectiveness and superiority of the proposed method, which can completely meet the actual needs of the process. • A Self-Learning Fuzzy Predictive Control Algorithm for Calculation of the Adjustment Variables of Cement Grinding Process. • Creating a 3-day cement strength prediction model based on step response and training it by difference data. • Proposing a multi-variable distributed fuzzy control algorithm and extracting fuzzy rules by Apriori algorithm. [ABSTRACT FROM AUTHOR]