Robotics and Computer-integrated Manufacturing最新文献

Reconfigurable manufacturing system is considered as a promising next-generation manufacturing paradigm. However, limited equipment and complex product processes add additional coupled scheduling problems, including resource allocation, batch processing and worker cooperation. Meanwhile, dynamic events bring uncertainty. Traditional scheduling methods are difficult to obtain good solutions quickly. To this end, this paper proposes a multi-agent deep reinforcement learning (DRL) based method for dynamic reconfigurable shop scheduling problem considering batch processing and worker cooperation to minimize the total tardiness cost. Specifically, a dual-agent DRL-based scheduling framework is first designed. Then, a multi-agent DRL-based training algorithm is developed, where two high-quality end-to-end action spaces are designed using rule adjustment, and an estimated tardiness cost driven reward function is proposed for order-level scheduling problem. Moreover, a multi-resource allocation heuristics is designed for the reasonable assignment of equipment and workers, and a batch processing rule is designed to determine the action of manufacturing cell based on workshop state. Finally, a strategy is proposed for handling new order arrivals, equipment breakdown and job reworks. Experimental results on 140 instances show that the proposed method is superior to scheduling rules, genetic programming, and two popular DRL-based methods, and can effectively deal with various disturbance events. Furthermore, a real-world assembly and debugging workshop case is studied to show that the proposed method is applicable to solve the complex reconfigurable shop scheduling problems.

The low stiffness of series robots limits their application in high-load precision manufacturing, such as automated fiber placement (AFP). This paper presents a stiffness optimization method to enhance the stiffness of plane-mobile robots in continuous fiber placement by simultaneously adjusting the robot's posture and the base position. A stiffness performance index suitable for evaluating the comprehensive stiffness of the robot during the AFP process is proposed, which is based on the fluctuation characteristics of the contact force in fiber placement. To maximize this index and the normal stiffness, the multi-objective particle swarm optimization algorithm (MOPSO) is used to solve the two-objective optimization model under multiple constraints. The constrained area of the mobile robot base corresponding to a given path point is determined by the fixed-height slice of the robot's reachable point cloud. A novel method combining global discrete solution and local continuous solution (GD-LC) is proposed to solve the model efficiently, which reduces the search dimension of the MOPSO algorithm. Experimental results from fiber placement on an aircraft mold show that the proposed method can significantly improve the stiffness performance of the AFP robot, and the force-induced deformation after continuous stiffness optimization is reduced by 70.01 % on average. The optimized laying quality further validates the engineering value of the proposed method.

Coverage path planning (CPP) has been widely studied due to its significant impact on the efficiency of automated surface quality inspection. However, these researches mostly concentrate on fixed-base visual robotic schemes, with limited focus on the widely utilized mobile-base schemes which require considerations of inherent constraints between stations (base positions) and viewpoints. Therefore, this article models a station-viewpoint joint coverage path planning problem and proposes a workflow to solve it. Within this workflow, firstly, a viewpoint selection genetic algorithm based on alternating evolution strategy is presented to optimize both the viewpoint quantity and view quality; secondly, a novel genetic algorithm is devised to accomplish joint assignment and sequence planning for stations and viewpoints. Several experimental studies are conducted to validate the effectiveness and efficiency of the proposed methods, and the proposed genetic algorithms exhibit notable superiorities compared to the benchmark methods in terms of viewpoint quantity, mean view quality, motion cost, and computational efficiency.

With the development of intelligent manufacturing and robotic technologies, the capability of grasping unknown objects in unstructured environments is becoming more prominent for robots with extensive applications. However, current robotic three-finger grasping studies only focus on grasp generation for single objects or scattered scenes, and suffer from high time expenditure to label grasp ground truth, making them incapable of predicting grasp poses for cluttered objects or generating large-scale datasets. To address such limitations, we first introduce a novel three-finger grasp representation with fewer prediction dimensions, which balances the training difficulty and representation accuracy to obtain efficient grasp performance. Based on this representation, we develop an auto-annotation pipeline and contribute a large-scale three-finger grasp dataset (TF-Grasp Dataset). Our dataset contains 222,720 RGB-D images with over 2 billion grasp annotations in cluttered scenes. In addition, we also propose a three-finger grasp pose detection network (TF-GPD), which detects globally while fine-tuning locally to predict high-quality collision-free grasps from a single-view point cloud. In sum, our work addresses the issue of high-quality collision-free three-finger grasp generation in cluttered scenes based on the proposed pipeline. Extensive comparative experiments show that our proposed methodology outperforms previous methods and improves the grasp quality and efficiency in clutters. The superior results in real-world robot grasping experiments not only prove the reliability of our grasp model but also pave the way for practical applications of three-finger grasping. Our dataset and source code will be released.

Five-axis machining, especially flank milling, is popular in machining thin-walled freeform surface parts with high energy consumption. Reducing the machining energy consumption is paramount for advancing green manufacturing. Therefore, this paper proposes an energy-efficient integration optimisation of cutting parameters and tool path with hierarchical reinforcement learning (HRL). Firstly, a novel multi-pass machining energy consumption model is developed with cutting and path parameters, based on which the integrated optimisation problem is modelled considering a dynamic workpiece deformation constraint. Secondly, HRL with a Soft Actor Critic agent (HSAC) decouples the model into two Markov Decision Processes at different timescales. The higher-layer plans cutting parameters for each pass on a macro timescale, while the micro-timescale lower-layer performs multiple tool path expansions with the planned cutting parameters, and provides feedback to the higher layer. By hierarchical optimisation and non-hierarchical interaction, the model is efficiently solved. Moreover, curriculum transfer learning is applied to expedite task completion of the lower layer, enhancing interaction efficiency between the two layers. Experiments show that, compared with two benchmarks, the proposed method improves machining energy consumption by 35.02 % and 30.92 %, and reduces machining time by 38.57 % and 27.17 %, providing a promising paradigm of green practices for thin-walled freeform parts and the broader manufacturing industry.

Additive and subtractive hybrid manufacturing (ASHM) refers to the hybrid manufacturing process where in-situ subtractive machining (SM) is introduced during additive manufacturing (AM). Its process characteristics dictate the necessity of planning multi-layer cutting paths in ASHM. Currently, the slice-based planning method cannot plan multi-axis cutting paths, and the machining accuracy is difficult to directly control. Meanwhile, the manual layering planning method is inefficient when dealing with complex models. Consequently, this paper presents an innovative automatic planning method for multi-layer, multi-axis, interference-free cutting paths with controllable precision in ASHM of composite enclosed cavity parts. To enhance the ASHM efficiency, criteria for the recognition of hybrid machining features (HMFs) have been defined to identify HMFs within the model. The identification of interference planes during cavity conversion has been achieved, and these interference planes are then utilized as the conversion planes for the ASHM process. Furthermore, a boundary-guided method is employed to automatically plan the overall cutting path for HMFs. According to the G-code standard, the overall cutting paths are then output to the corresponding cutting path file within the height interval of the conversion planes. Through practical machining, it has been demonstrated that the proposed method can significantly enhance the efficiency and automation of the data preparation process in ASHM, while also improving the surface quality and dimensional accuracy of the AM part.

Current research on the disassembly line balancing problem ignores the influence of non-disassemblability of components. And this problem can lead to failure of the disassembly task, which can seriously affect the disassembly efficiency. This study integrates destructive operation into the human-robot disassembly line while considering noise. First, a mixed integer programming model is established for human-robot hybrid partial destructive disassembly line balancing problem to accurately obtain the number of stations, smoothness index, costs and negative impact of noise pollution on workers. Then, an improved grey wolf optimization algorithm is proposed for the NP-hard characteristic of problem. A three-layer encoding and two-stage decoding strategy is designed to constrain the uniqueness of the solution, considering the noise constraints, and the different disassembly times of the human-robot. A disturbance factor is also designed to prevent local optimality, which enhances the performance of the proposed algorithm. Different cases are also used to verify the correctness and superiority of the proposed method. Finally, an engine case is used to validate the practicality of the proposed method. The results of the comparison of the different disassembly schemes show that: (1) The proposed algorithm outperforms the three classical Swarm Intelligence methods and other eleven algorithms in the disassembly line balancing problem. (2) The human-robot hybrid partial destructive disassembly line can effectively avoid the problem of task failure, and the smoothing index is reduced by 12.27 % compared with the original scheme. Disassembly costs increased by 1.28 %, but this was minimal compared to line-wide smooth running and worker health. (3) The human-robot hybrid disassembly line is more appropriate to solve the actual production process compared to worker disassembly and robot disassembly, and has a greater advantage in solving the actual disassembly line balance problem.

In recent years, robotic machining has become one of the most important paradigms for the machining of large and complex parts due to the advantages of large workspaces and flexible configurations. However, different configurations will correspond to very different system performances, influenced by the position-dependent properties. Therefore, the configuration optimization of robotic machining system is the key to ensure the quality of robotic operation. In response to the fact that little attention has been paid in current research to the effect of mapping model distribution differences on the optimization results, a sparse knowledge embedded configuration optimization method for robotic machining systems toward improving machining quality is proposed. The knowledge of theoretical model-based optimization in terms of stage, density and redundancy is embedded into high-fidelity data by three steps sparse and real measurement. Pre-training and domain adaptation fine-tuning strategies are used to reconstruct the real mapping model accurately. The reconstructed mapping model is re-optimized to obtain a more accurate system configuration. The effectiveness of the proposed method is verified by machining experiments on space segment parts. The proposed method reduces the absolute position error and machining error by 48.67 % and 28.73 %, respectively, compared to the current common theoretical model-based optimization. This is significant for more accurate and reliable robot system optimization. Furthermore, this work confirms the influence of mapping model distribution differences on the optimization effect, providing a new and effective perspective for subsequent research on the optimization of robotic machining system configurations.

In this paper, we present a new computer-controlled weaving technology that enables the fabrication of woven structures in the shape of given 3D surfaces by using threads in non-traditional materials with high bending-stiffness, allowing for multiple applications with the resultant woven fabrics. A new weaving machine and a new manufacturing process are developed to realize the function of 3D surface weaving by the principle of short-row shaping. A computational solution is investigated to convert input 3D freeform surfaces into the corresponding weaving operations (indicated as W-code) to guide the operation of this system. A variety of examples using cotton threads, conductive threads and optical fibers are fabricated by our prototype system to demonstrate its functionality.

Robot skill learning is one of the international advanced directions in the field of robot-based intelligent manufacturing, which makes it possible for robots to learn and operate autonomously in complex real-world environments. In this paper, we propose a multitasking-oriented robot skill learning framework named MT-RSL to improve the efficiency and robustness of multi-task robot skill learning in complex real-world environments, and present the detailed design steps of three key sub-modules included in MT-RSL, namely, sub-task segmentation module, robot skill learning module, and robot configuration optimization module. Firstly, we design a novel sub-task segmentation module based on a coarse-to-fine sub-task segmentation (CF-STS) strategy, in which the Finite State Machine (FSM) is used to analyze complex robot behaviors to obtain a coarse robot sub-task sequence, and the Beta Process Autoregressive Hidden Markov Model (BP-AR-HMM) is used to establish the connection and dependence between multiple demonstration trajectories and encode these trajectories, so as to obtain a finer robot action sequence. Secondly, we extend the basic DMPs system to a continuous dynamic movement primitives (CDMPs) system to construct a novel robot skill learning module, which improves the efficiency of the robot to learn skills and perform actions by orderly coordinating sub-parts such as the activation signals, motion actuator, DMPs-based learning module, and robot configuration optimization module. Then, we design a novel robot configuration optimization module, which introduces the velocity directional manipulability measure (VDM) as the evaluation index of robot kinematic performance to establish the robot configuration optimization model, and proposes an improved probabilistic adaptive particle swarm optimization (Pro-APSO) algorithm to solve this optimization model, so as to obtain the optimal robot configuration. Finally, we develop an experimental testing platform based on the Robot Operating System (ROS) and conduct a series of prototype experiments in complex real-world scenarios. The experimental results demonstrate that our proposed MT-RSL can significantly improve the effectiveness and robustness of multi-task robot skill learning, and can outperform existing robot skill learning methods in terms of both learning efficiency, VDM, and success rate.