In light of the growth in data availability, the manufacturing industry is experiencing a growing change in its needs and shape, which necessitates the use of more efficient data-driven methodologies in real-time production scheduling. Conventional dynamic scheduling approaches, designed for non-dynamic problem sizes, struggle in dealing with the inherent volatility and complexity of contemporary production scheduling scenarios. As an attempt to meet the demands of near real-time decision-making mechanisms on the shop floor, this study explores variations of QMIX, including local QMIX (LQMIX), where separate instances are employed for specific tasks, and gated QMIX (GQMIX), which utilizes specific agents for tasks while employing a central mixing network. Scheduling systems may not satisfy the recent requirements, emphasizing the need for more adaptive systems. Utilizing QMIX, manufacturing operations can be streamlined by integrating the collaborative synergy of multiple agents. Reinforcement learning models are trained using QMIX and benchmarked against heuristic dispatch strategies such as Shortest Path First as one of the most popular method used in the community. The experimental findings highlight the effectiveness of QMIX, in particular the original version, in tackling the challenges of dynamic scheduling within the manufacturing domain. QMIX exhibits superior performance compared to alternative algorithms and heuristic dispatching methods in specific contexts. Nonetheless, the study underscores the imperative of striking a balance between adaptability and specialization.

For many years, universal robots have been exten-sively popular across a wide range of industries to accommodate a broad range of manufacturing requirements. The complexity of manufacturing environments has resulted in a growing demand for automated programming approaches for robotic systems, where artificial intelligence approaches have recently proven effective. Despite recent advancements, the development of AI-driven controllers for universal robots performing complex, multi-step tasks continues to face several challenges, including stability issues. This paper aims to address these challenges by leveraging reinforcement learning to automate robot programming, with a particular focus on ensuring the stability of robot movements during multi-step robot positioning for inspection purposes. Our research formulates the multi-step robot positioning task as a reinforcement learning problem and develops a reward function to account for the robots' stability at the checkpoints. We conducted experiments using four different RL methods, namely PPO, TRPO, SAC, and TD3. Our findings indicate that TRPO outperforms the other methods, converging to an optimal controller. This study contributes to the field of robotics by providing a robust approach to enhancing the stability and efficiency of robot programming in complex manufacturing environments.

Object detection is a task in computer vision that involves detecting instances of visual objects of a particular class in digital images. Numerous computer vision tasks highly depend on object detection such as instance segmentation, image captioning and object tracking. A major purpose of object detection is to develop computational models that provide inputs crucial to computer vision applications. Convolutional Neural Networks (CNNs) have recently become popular due to their key roles in enabling object detection. However, the performance of CNNs is largely dependent upon the quality and quantity of training datasets, which are often difficult to obtain in real-world applications. In order to ensure the robustness of such models, it is vital that training instances are provided under various randomized conditions. These conditions are typically a combination of a variety of factors, including lighting conditions, object location, the presence of multiple objects in the scene, varieties of backgrounds, and the angle of the camera. In particular, companies, depending on their applications (such as fault detection, anomaly detection, condition monitoring, predictive quality and so on), require specialized models for their custom products and so always face difficulties in providing a large number of randomized conditioned instances of their objects. The primary reason is that the process of capturing randomized conditioned images of real objects is usually costly, time-consuming, and challenging in practice. Due to the efficiency gained so far via the use of synthetic data for training such systems, synthetic data has recently attracted considerable attention. This paper presents an end-to-end synthetic data generation method for building a robust object detection model for customized products using NVIDIA Omniverse and CNNs. In this paper, we demonstrate and evaluate our contribution to the modeling of chess pieces, where a total accuracy of 98.8 % was obtained.