Model predictive control (MPC) has been widely employed in autonomous driving control, offering advantages including rolling optimization. However, the performance of MPC can be significantly compromised due to the inaccurate system model of the vehicle. In response, a novel MPC model optimization algorithm based on BiLSTM has been proposed to address the issue of inaccurate system models. Firstly, MPC-based pathtracking data is collected under varying speeds, paths, and wheelbases. A BiLSTM network is then trained using this dataset and integrated with a MPC controller. This integration allows for the update of the wheelbase parameters of the kinematic model, thereby improving the accuracy of the vehicle prediction model. The proposed method is validated through simulation experiments. The results demonstrate that the path tracking accuracy of the MPC combined with a BiLSTM method is higher than that of MPC and linear quadratic regulator (LQR). Concurrently, this approach yields a considerable enhancement in path tracking precision, with an improvement of 37.53%.
Map fusion is a key challenge in multi-robot SLAM systems, with feature-based methods often leading to mismatches when handling maps of different resolutions. This paper introduces a density clustering-based map fusion method that reduces mismatches by clustering features before matching them. Experiments on public datasets demonstrate that the proposed method effectively merges grid maps of varying resolutions, offering improved efficiency and accuracy compared to traditional feature-based methods.
The path planning of Unmanned Underwater Vehicle (UUV) is a crucial aspect of their operation in underwater environments, Meta-heuristic algorithms are extensively utilized for addressing UUV path planning problems. To address the limitations of the traditional dung beetle optimization algorithm (DBO), including inadequate convergence speed and precision in two-dimensional UUV path planning, and its propensity for local optima, an improved dung beetle optimization algorithm (IDBO) is introduced which employing a suite of refinement strategies. Furthermore, the solution capability of the IDBO is validated through the CEC2017 test suite and two-dimensional raster maps that replicate actual underwater environments. The simulation results demonstrate the robust problem-solving capacity of the IDBO, applicable to both benchmark functions and real-world scenarios, affirming the efficacy of the enhancement strategies in practical applications.
Mobile robot machining system offers an effective solution for the integrated machining of large-scale components. However, the lower absolute positioning accuracy of industrial robots poses a significant challenge to achieving high-precision manufacturing of such components. Identifying robot kinematic parameters and compensating for motion errors represent effective approaches to enhancing the absolute positioning accuracy of robots. In this paper, we propose a two-step method for identifying robot kinematic parameters, which combines the LASSO algorithm with the IPSO algorithm, aiming to improve the absolute positioning accuracy of robotic machining systems for large-scale components. Initially, values of robot kinematic parameter deviations are obtained by applying the LASSO algorithm. These values then serve as initial particles for the IPSO algorithm, enabling more efficient and accurate kinematic parameter identification. Subsequently, the kinematic parameters of the robot machining system are efficiently calibrated. Experimental results demonstrate that the maximum absolute positioning error is improved by over 90% before calibration, demonstrating the validity of the proposed method in enhancing the absolute positioning accuracy of industrial robots.
By analyzing some disadvantages of existing spraying robot configuration, such as the lack of space and the destruction of the coating in the complex cavity surface, aiming at the special needs of the complex cavity surface spray, starting from the efficiency and spray quality of industrial production, a new type of six-degree-of-freedom robot is proposed. By combining the Modified Denavit-Hartenberg method and the anti -view angle algorithm, the robot’s kinematics model was established, and the kinematics solution and workspace were performed to verify the performance of the robot space. Taking the aircraft intake as an example, by simulating the machining trajectory of the aircraft intake, robotics simulation is performed to obtain the dynamic parameters of the axis joints, which provides a basis for subsequent motor selection and strength check. Through workspace solution and kinematics simulation verification, it proves that the agency not only can meet the high precision and high efficiency requirements of the complex cavity surface, but also shows the potential of robotic technology in improving production efficiency, ensuring the quality of spraying and achieving automated production. It provides new ideas and solutions for the development of spraying robot technology and its promotion in industrial applications.
Aiming at the problems of low efficiency, high danger and high cost of manual disc cutter changing during the construction of shield structure method, an automated disc cutter changing program based on robot operation is proposed. Firstly, by analyzing the operating environment and functional requirements of the disc cutter changing robot, a new 8-degrees-of-freedom foldable disc cutter changing robot structure is designed. Then, based on the ANSYS Workbench simulation platform, static calculation and simulation are performed to verify that the strength and stiffness of the whole structure of the robot meet the design requirements; finally, the kinematic modeling of the robot is performed by the D-H method, and the workspace calculations are performed by the Monte Carlo pseudo-random method. The results show that the disc cutter changer robot can meet the functional requirements, motion requirements and rigidity requirements for changing the hob under the premise of adapting to the narrow space inside the TBM, and can realize the replacement of 90.2% of the disc cutter of the whole machine.
Large and complex components are the core parts of major equipment. In the in-situ machining of these components, the precise alignment of cutting tools and part features is very important and also very challenging, and accurate measurement is the foundation. However, the reflective metal surfaces of these large components present significant challenges for the detection and measurement of visual markers on their surfaces. This paper presents a method for the detection and the three-dimensional reconstruction of markers on high-reflective components. The proposed method integrates the DEtection TRansformer (DETR) model and employs a data augmentation strategy to enhance model generalization. Segmentation of the regions of interest (ROI) of markers is employed to mitigate the issue of high reflectivity on surfaces. An enhanced ellipse detection method is developed that integrates multi-channel information from RGB images with a weighted grayscale image, leading to a significant improvement in the detection rate of visual markers. The effectiveness of the method is validated through comparative experiments with a laser tracker. The measurement experiment results show that the maximum measurement error is 0.0523 mm, and the average error is 0.0363 mm. The accuracy of the proposed method can meet the actual requirements, and the problem of high-reflection is solved. The study presents a feasible solution for marker detection and measurement in in-situ manufacturing of large and complex components.
Multi-robot cooperative exploration can reduce the time required for exploration tasks in unknown environments and improve system efficiency. This paper improves two aspects of the frontier-based multi-robot exploration method. Firstly, we propose a two-stage frontier clustering algorithm. First-stage clustering aggregates frontiers into new frontiers based on continuity, significantly reducing their quantity, and then a GriTDBSCAN clustering is performed to effectively remove small and dense frontier tasks, thereby reducing the number of points to be evaluated and improving detection efficiency. Additionally, the number of obstacles between robots and task points is incorporated into the utility value calculation to avoid forming detection islands. The proposed method was evaluated by simulating four contrasting methods in two different environments. The results clearly indicate that the proposed method outperforms the others, yielding the best detection results.
Aiming at the problems in the design of traditional involute internal gear, such as the limitation of teeth number, the inflexibility of the profile shift coefficient and the low contact ratio, a method of internal gear with high contact ratio (HCR) based on predefined circular arc meshing line is provided. In order to realize the HCR, the circular arc connecting the intersection point of the inner gear and the outer gear tooth tip circle to the pitch point is used as the meshing line. According to the principle of gear meshing theory, firstly, an equation expression for the conjugate tooth profile curve that satisfies the circular arc meshing line is established, and the dedendum profile curve conjugated with the addendum profile curve is constructed. Then, the dedendum profile curve of the internal gear is modified to avoid simultaneous contact between two points. Finally, the tooth dedendum profiles of the internal and external meshing gears were designed. Compared with traditional involute gears, gears with HCR have greater coincidence and therefore have greater load-bearing capacity. The finite element analysis software ABAQUS is used to simulate the HCR gear meshing process. The simulation results show that, after the modification, the problem of sudden change in meshing force is improved, and the relative sliding rate of meshing in and out is reduced from -1.176/0.593 to -0.823/0.509. The maximum power loss during the meshing process is reduced from 930w to 850w, and the meshing efficiency is increased from 99.267% to 99.333%.
In the construction process of shield machine, a large number of cutters need to be replaced, and the traditional method of manual cutter replacement has a complicated replacement process, the cutter changing time is long, the risk is high, and the cost is high. Therefore, more and more scholars at home and abroad have begun to study the “machine cutter changing”. In view of the large load, high terminal positioning accuracy and small activity space of the cutter-changing robot, the electro-hydraulic servo control system of the cutter-changing robot was designed to realize the tool change function of the robot, the mathematical model of the electro-hydraulic servo system was established, the working principle of the fuzzy PID controller was analyzed, the traditional PID controller and the fuzzy PID controller were built for MATLAB simulation, and the simulation effects of the traditional PID controller and the fuzzy PID controller were compared. The simulation results show that both traditional PID control and fuzzy PID control can improve the system performance, and the effect of fuzzy PID control is better than that of traditional PID control.
This paper presents an actuator motor control scheme of the electric mover for the non-powered trailer. The actuator motor of the electric mover drives the traction roller in the forward and reverse direction to attach and detach the traction roller to the trailer wheel. Because the trailer motion depends on the wheel motion transferred from the mover roller, the contacting strength of the mover roller is very important to keep the constant traction power transmission. This paper presents sensorless speed estimation and non-linear disturbance observer to improve the constant traction power transmission performance. The proposed actuator motor control scheme determines the braking position by the estimated disturbance torque to press the trailer wheel. To estimate the disturbance torque of the actuator motor, the actual motor speed has to be detected. The sensorless estimation scheme obtains the actuator motor speed in the proposed method. Then, the estimated actuator motor speed is used to observe the nonlinear disturbance torque of the roller to the trailer wheel. The observed disturbance torque shows the starting mechanical friction and the reflected disturbance torque from the trailer tire well. The observed disturbance torque is used to determine the stop position of the mover roller to keep the constant friction between the roller and the trailer wheel.
Endoscopic inspection technology is mainly used in regular maintenance for aeroengines. However, there are some inner components that are hard to be observed by this means, such as stator blades, which leaves many potential safety troubles. Recently there have been some robots emerging for aeroengine maintenance and inspection, which have undergone significant changes compared to current methods. But at present they are still unable to achieve in engineering applications for their immaturities. To increase the possibility of problem solving, this paper provided a summary of them, and presented a novel approach against their shortcomings, which involves an endoscopic inspection robot. According to the general design of this robot, the typical structures of both main parts, the snakelike arm and the passive blade grasper were given.
The integration of CNNs and transformers is a challenging task that requires efficiently incorporating of both architectures. Many previous studies neglect the significance of vanilla convolutional features and their channel information. In this paper, we propose a new model named Channel Transformer (ChanT), which includes a multi-channel self-attention module that focuses on channel information at low resolution and spatial information at high resolution. ChanT supports all CNN architectures due to its compatibility with convolution. Experiments on object detection tasks demonstrate its ability to generalize on large datasets and transfer to small datasets. ChanT also bridges the gap between CNNs and Transformers on small datasets. Results show that Yolov5 outperforms ChanT by less than +1.5 AP on PASCAL VOC, and ChanT-L trained from scratch outperforms Res50 and Swin-T pretrained on ImageNet-1k by +4.6 AP and +0.4 AP on COCO 2017, respectively.
Point clouds can lead to the problem of losing key geometric features during the simplification process. A point cloud simplification algorithm based on classification simplification strategy is proposed for this purpose. Firstly, the point cloud is simplified to achieve a balance between simplification and geometric features. Then, the improved region growing algorithm is used to accurately segment the point cloud and simplify the sampling of different degrees of voxels. The integrity of the key geometric features is ensured, and the simplified accuracy is evaluated by the RMSE. Finally, the measurement verification is carried out on the circular hole of the thin-walled part. The results show that the proposed method is superior to the traditional voxel down sampling, and can deal with complex multi-feature point cloud models. It can effectively solve the problem of missing feature points in the traditional simplification process and meet the measurement dimensional accuracy requirements of industrial digital manufacturing.
To meet the needs of high-precision and highefficiency measurement of large-scale aerospace components, a new measurement and registration technology with large-field-of-view is proposed in this paper, which is used to measure large-scale and weak-textured complex components. The robotic arm drives a small-field-of-view, high-spatial-resolution binocular fringe projection local measurement system, which acquires highprecision point cloud data based on the principle of heterodyne multi-frequency phase shift and high dynamic measurement. Based on the registration method of phase matching and global marker points, the local sensors are tracked and located by the global sensors with large-field-of-view and low-spatial-resolution. The laser tracker and the global sensor are used to jointly calibrate the global marker points on the surface of the turntable to realize the registration of all areas of large components. An automatic measurement method of the geometric dimensions of components by processing the point cloud to extract geometric feature and a virtual assembly and assembly quality evaluation method based on the point cloud data of components to optimize the assembly sequence and recognize the assembly interference of components are proposed in this paper.
In this paper, the electromagnetic compatibility requirements of the vehicle indirect vision devices are described, and a radiation immunity verification test method by using an excitation light source is introduced. The layout and principles of the time delay test system of the camera-monitor system (CMS) are described. And the influence factors of the excitation light source system on the test delay of the CMS system are studied, the influence degree and reason analysis of different factors are verified by the test data.
In order to monitor the wear amount and wear state of the disc cutter ring in real time during the tunneling process of TBM (Tunnel Boring Machine), an on-line monitoring system was designed to collect the wear amount and rotational speed and identify the wear state through BP neural network. The system uses an eddy current sensor to convert the wear between the disc cutter ring and the sensor into a voltage signal, which is transmitted to the host computer after A/D conversion. The BP neural network model is used in the host computer to identify the wear state of the disc cutter ring. After testing and verifying on a 1:2 scale experimental bench, the results show that the monitoring system can accurately detect the wear amount of the disc cutter ring and identify the wear state of the disc cutter. The determination coefficient of the test sample is 0.92873, and the root mean square error is 0.064975, which can realize the online monitoring of the wear of the disc cutter ring.
This paper presents a novel path planning algorithm for a picking robotic arm in a multi-obstacle environment, based on deep reinforcement learning. The proposed method introduces a new state representation technique that accurately captures the real-time state information of the robotic arm and multiple obstacles using a finite-dimensional representation. This state is represented by the direction vector of the nearest distance to the nearest obstacle around each axis of the robot arm, the real-time angle of the robotic arm, the three-dimensional coordinates of the picking target, and the three-dimensional coordinates of the end effector. The effectiveness of this method is validated through tests in a simulation environment.
High-precision line laser scanning measurement technology has become an important means of surface inspection for key aerospace components. To improve detection efficiency, the million to ten million point clouds acquired by scanning must be simplified. Due to the complexity of the surface features (stepped planes, holes, etc.), the traditional point cloud simplification methods can hardly consider the simplification rate and the completeness of key features simultaneously. To meet the needs of matching point clouds with 3D models, this paper develops a point cloud simplification method that can preserve the boundaries of the point cloud based on Intrinsic Shape Signature (ISS) key points. Firstly, the point cloud boundary is extracted. Then, the feature points and the ISS key points are extracted. Finally, all point sets obtained are merged, and duplicate points are removed. This paper builds a 3D measurement system with a robot-mounted line laser scanner and a linear displacement stage. The scanning experiments are conducted using three types of brackets. Comparing the developed method in this paper with curvature-based grading methods and point by point forward methods, the results show that our method outperforms other methods
Autonomous Underwater Vehicle (AUV), as an unmanned exploration equipment, is extensively employed in both military and civilian sectors, bearing significant importance for exploitation of marine resources. As a core component, an efficient and reliable propulsion system is crucial to the overall performance and operational safety of the AUV. This paper first designs the exterior of the AUV based on the operating conditions, and utilizes ANSYS FLUENT to perform hydrodynamic analysis, thereby determining navigation resistance. Subsequently, on the basis of propeller propulsion, a crank-slider mechanism is adopted to convert reciprocating motion into continuous rotation. The reciprocating push rod motor is submerged in oil, and the flexible oil bladder separates the motor from the mechanical transmission components. The deformation characteristics of the flexible bladder under different displacement conditions are simulated and analyzed using COMSOL to further validate the rationale behind the reciprocating push rod motor propelling the propeller system. Finally, the processing and assembly of the AUV prototype are completed, and relevant underwater tests are conducted.
