In the field of robotics, particularly in human–robot interaction, rapid and precise torque control is crucial. This study presents a closed-loop iterative optimized (CIO) approach designed to improve current tracking in servo systems, positioning it as a feasible alternative to traditional proportional-integral controllers. The proposed method innovatively combines a fractional-order proportional-integral-derivative (FOPID) controller with linear matrix inequality (LMI) technique and genetic algorithm in a closed-loop iterative setting. This integration not only boosts the FOPID controller's response speed, but also leverages the robustness of LMI, enhancing stability under various operating conditions. The inclusion of heuristic algorithms in this system enables a more dynamic and efficient search for optimal control parameters, aligning the controller's performance with the complex requirements of contemporary robotics. Experimental data underscore that this approach significantly increases torque response speed and the control system's bandwidth, vital for satisfying the dynamic needs of current robotic applications.

Panoramic image generation enhances visual representation and facilitates immersive experiences across various fields. However, Existing models readily gives rise to error accumulation and loop closure. To overcome these limitations, we introduce PGDM, a diffusion-based framework that facilitates the concurrent generation of visually cohesive panoramic images. In order to achieve modality alignment, PGDM adopts the vision-language model as a Multimodal Module for extracting visually representative text-aligned features. Additionally, we design the Multi-View Attention with Camera Pose mechanism between each UNet layer to enhance the coherency of panoramic images. Our extensive experimental results demonstrate that PGDM achieves the state-of-the-art performance while simultaneously maintaining diversity and consistency across multiple view images.

Diffusion models hold significant appeal within the realm of synthetic media generation and demonstrate exceptional performance in personalized human image generation.However, the efficiency of the existing approaches is hindered by their limited capacity to handle images captured under different exposure conditions as the models tend to incorrectly attribute the illumination of an individual's facial region to their complexion.In addition, previous methodologies encounter challenges in the realm of posture transfer due to the fact that their image encoders primarily emphasize the extraction of character identity.We introduces ExpoGenius, a framework that facilitates efficient and personalized human image generation for exposure variation and pose transfer, all accomplished without the need for fine-tuning.ExpoGenius employs Multi-Grained Identity Feature to extract and combine facial features that are not affected by varying lighting conditions, enabling the accurate representation of facial attributes in various diverse exposure levels.To facilitate the transfer of pose , we propose Pose Feature Injection to inject pose feature into our improved diffusion model with identity feature and textual embeddings.In scenarios where facial images occupy a relatively small proportion within an overall image, the accuracy of identity extraction may encounter challenges.ExpoGenius presents Adaptive Facial Loss which effectively enhances the accuracy of identity extraction.Our research has substantiated the exceptional effectiveness of ExpoGenius in simultaneously preserving identity and posture in personalized human image generation tasks.

G-protein coupled receptors (GPCRs), crucial in various diseases, are targeted of over 40% of approved drugs. However, the reliable acquisition of experimental GPCRs structures is hindered by their lipid-embedded conformations. Traditional protein-ligand interaction models falter in GPCR-drug interactions, caused by limited and low-quality structures. Generalized models, trained on soluble protein-ligand pairs, are also inadequate. To address these issues, we developed two models, DeepGPCR_BC for binary classification and DeepGPCR_RG for affinity prediction. These models use non-structural GPCR-ligand interaction data, leveraging graph convolutional networks and mol2vec techniques to represent binding pockets and ligands as graphs. This approach significantly speeds up predictions while preserving critical physical-chemical and spatial information. In independent tests, DeepGPCR_BC surpassed Autodock Vina and Schrödinger Dock with an area under the curve of 0.72, accuracy of 0.68 and true positive rate of 0.73, whereas DeepGPCR_RG demonstrated a Pearson correlation of 0.39 and root mean squared error of 1.34. We applied these models to screen drug candidates for GPR35 (Q9HC97), yielding promising results with three (F545-1970, K297-0698, S948-0241) out of eight candidates. Furthermore, we also successfully obtained six active inhibitors for GLP-1R. Our GPCR-specific models pave the way for efficient and accurate large-scale virtual screening, potentially revolutionizing drug discovery in the GPCR field.

Accurate and reliable segmentation of colorectal tumors and surrounding colorectal tissues on 3D magnetic resonance images has critical importance in preoperative prediction, staging, and radiotherapy. Previous works simply combine multilevel features without aggregating representative semantic information and without compensating for the loss of spatial information caused by down-sampling. Therefore, they are vulnerable to noise from complex backgrounds and suffer from misclassification and target incompleteness-related failures. In this paper, we address these limitations with a novel adaptive lesion-aware attention network (ALA-Net) which explicitly integrates useful contextual information with spatial details and captures richer feature dependencies based on 3D attention mechanisms. The model comprises two parallel encoding paths. One of these is designed to explore global contextual features and enlarge the receptive field using a recurrent strategy. The other captures sharper object boundaries and the details of small objects that are lost in repeated down-sampling layers. Our lesion-aware attention module adaptively captures long-range semantic dependencies and highlights the most discriminative features, improving semantic consistency and completeness. Furthermore, we introduce a prediction aggregation module to combine multiscale feature maps and to further filter out irrelevant information for precise voxel-wise prediction. Experimental results show that ALA-Net outperforms state-of-the-art methods and inherently generalizes well to other 3D medical images segmentation tasks, providing multiple benefits in terms of target completeness, reduction of false positives, and accurate detection of ambiguous lesion regions.

Named entity recognition is one of the most fundamental problems in knowledge graph. In biomedical field, labeling high-quality biomedical entities requires plenty of linguistic knowledge due to abbreviation and specificity. Using dictionary is the simplest way for labeling, but it is difficult to obtain a versatile dictionary and usually a dictionary for one corpus is not suitable for another corpus due to bad transferability. Current mainstream recognition methods require lots of manpower, which is time-consuming and laborious. To handle this challenge, we present a novel approach to automatically recognize new biomedical entities. First, we use a small number of manually labeled biomedical entities as seeds to label some biomedical texts and learn their features autonomously. Then by using a tagger based on supervised learning and an instance selector based on reinforcement learning, we iteratively generate new biomedical entities. Experiment results demonstrate that our method can deal with biomedical named entity recognition and obtain significant performances in both English and Chinese biomedical datasets.

Named entity recognition (NER) is an essential task in information extraction and knowledge graph construction. Numerous traditional methods for NER are sentence-level, which only utilize all the information in one sentence and predict entities inconsistently sometimes. Recently, researchers observe that the relationship between sentences in a document is helpful for the NER task, therefore they consider to conduct documentlevel approaches which capture context information within a document. Though these methods are effective and have been widely used, they ignore correlations between sentences in different documents. To make full use of the cross-document context information, we design an attention mechanism to model the semantic association between occurrences of the same word in different documents. In addition, we find that it is extremely helpful to add a multi-classification auxiliary task, which focuses on coarse-grained entity information. The multi-objective optimization is implemented by weighted summation and an autonomous approach using the homoscedastic uncertainty of each task. Extensive experiments in different datasets show that our approach is effective, and our models achieve better results than the sentence-level and document-level NER approaches.