With the development of artificial intelligence technology and popularity intelligent production projects, inspection systems have gradually become a hot topic in industrial field. As fundamental problem field computer vision, how to achieve object detection industry while taking into account accuracy real-time is an important challenge systems. The defects on steel surfaces application industry. Correct fast surface can greatly improve productivity product quality. To this end, paper...

Tool wear is a key factor in the machining process, which affects tool life and quality of machined work piece. Therefore, it crucial to monitor diagnose condition. An improved CaAt-ResNet-1d model for multi-sensor diagnosis was proposed. The ResNet18 structure based on one-dimensional convolutional neural network adopted make basic architecture. more suitable feature extraction time series data. Add channel attention mechanism CaAt1 residual block CaAt5 automatically learns features...

Visual servoing is widely used in the peg-in-hole assembly due to uncertainty of pose. Humans can easily align peg with hole according key visual points/edges. By imitating human behavior, we propose P2HNet, a learning-based neural network that directly extract desired landmarks for servoing. To avoid collecting and annotating large number real images training, built virtual scene generate many synthetic data transfer learning. A multi-modal strategy then introduced combine image-based...

Redundant manipulators are widely used in fields such as human-robot collaboration due to their good flexibility. To ensure efficiency and safety, the manipulator is required avoid obstacles while tracking a desired trajectory many tasks. Conventional methods for obstacle avoidance of redundant may encounter joint singularity or exceed position limits trajectory. By integrating deep reinforcement learning into gradient projection method, reactive method proposed. We establish general DRL...

On-orbit operation tasks require the space robot to work in an unstructured dynamic environment, where end-effector’s trajectory and obstacle avoidance need be guaranteed simultaneously. To ensure completability safety of tasks, this paper proposes a new obstacle-avoidance motion planning method for redundant robots via reinforcement learning (RL). First, framework, which combines RL with null-space robots, is proposed according decomposition joint motion. Second, model constructed, agent’s...

This paper presents an efficient path planning method for the lunar rover to improve autonomy and exploration ability in complex unstructured surface environment. Firstly, safe zone rover’s motion is defined, based on which a detecting point selection strategy proposed choose target positions that meet constraints. Secondly, improved sampling-based get efficiently. Thirdly, map extension continually varying environment included update roadmap, takes advantage of historical information....

Probabilistic roadmap (PRM) can effectively solve the path planning problem in environment with high-dimension and complex constraints, but has limitations of low quality efficiency narrow channel dynamic environment. In order to improve applicability characteristics PRM, a for mobile robot Multi-dimensional based on PRM blended potential field is proposed this paper. It contains three main parts: Firstly, set workspace, then number adaptive sampling points area-division are carried out...

Abstract This study examines the impact of stock market liberalisation on managerial environmental, social and governance (ESG) learning from prices. Using a quasi‐natural experiment, specifically Shanghai‐Hong Kong Stock Connect (SHHKC) Shenzhen‐Hong (SZHKC), we find that enhances firms' ESG expenditure sensitivity to prices, implying managers extract greater amounts information Additionally, mechanism test shows influences by enabling prices incorporate foreign investors' private at both...

In this paper, a dynamic allocation approach with task sequence modification to accommodate on-orbit tasks and various space robot performances is proposed. An Evaluation for the efficiency of tasks, including workspace motion performance robot, built using greedy auction algorithm. To further increase algorithm's running speed, an adjustment be assigned introduced in order reduce search solution assignment algorithm, thereby reducing algorithm calculation time. The simulation results show...

Aiming at the redundant manipulator operation task that needs to ensure end-effector trajectory tracking as much possible in dynamic obstacle scene, a loose null-space avoidance (LNOA) method based on reinforcement learning (RL) is proposed. Firstly, joint motion decomposed into and motion, latter further slack motion; this basis, LNOA framework for designed. Secondly, RL introduced learn generation strategy, so generate component autonomously, which then combined with realize maintenance...

As professional football stadiums continue to grow in popularity worldwide, fans are able watch the game closer proximity, but design of shorten distance between and playing field also exacerbates impact home advantage on referee’s decision call a penalty. Studies have confirmed existence found that experienced referees can reduce this interference, neural mechanisms behind phenomenon not been adequately investigated. In study, we designed soccer referee making task based effect scenario...

5G is an essential platform for realizing industrial upgrading. Various industries including manufacturing have established a wide range of 5G-enabled use cases. Many the networks used in these cases are private network or created specifically by business customers. Thus, expected to carry significant future traffic. This paper introduces China Telecom's technology networking mode and provides architecture proposals end-to-end customized solution. Afterward, this considers critical aspects,...

The main challenge of the decentralized control method for reconfigurable modular robots lies in handling interconnection terms and joint friction torque. In this paper, an adaptive terminal sliding mode is proposed trajectory tracking problem robots. Firstly, robot dynamics equations are rewritten form equations, a radial basis neural network used to fit approximation local information term; terms, its bound function transformed into nonlinear compensation term designed compensate it; LuGre...

A two-time scale trajectory tracking and vibration suppression controller is proposed for non-planar flexible space manipulator with uncertain dynamic parameters. Firstly, the model of considering multi-dimensional flexibility link built. Secondly, decomposed by singular perturbation method into a slow subsystem fast subsystem. Then, two sub-controllers adaptation parameter uncertainty are designed respectively combined controller. Finally, numerical simulation carried out to verify validity method.