http://yetl.yabesh.ir/yetl1/handle/yetl/19045 2026-04-23T12:46:19Z http://yetl.yabesh.ir/yetl1/handle/yetl/4310978 Pre-Training of a Large Robotic Design Model Li, Guannan; Song, Zehong; Huo, Xinming; Sun, Tao Robotic design is a complex, multiparametric, and nonlinear process characterized by the intricate mapping between design requirements and solutions. Traditional methods often face limitations due to sequential workflows and human-induced biases, while conventional artificial intelligence models struggle to generalize across diverse design tasks. To address these challenges, we propose a novel cross-modal pretraining framework: robotic language-graph pretraining (R-CLGP). This framework bridges unstructured natural language requirements with structured robot designs, leveraging large-scale datasets for pretraining and flexible adaptation to various design requirements. The R-CLGP model utilizes a graph-based representation method that captures both non-Euclidean and Euclidean features and contrastive learning to enhance the mapping of textual requirements to robot topologies, significantly improving design efficiency and enabling intuitive design interaction. Through use cases such as requirement–topology retrieval, topology–topology retrieval, and performance prediction, the framework demonstrates its ability to streamline robotic design by minimizing manual intervention and improving scalability. This work not only advances methodologies in robotic design but also offers a transformative and adaptable framework for broader applications in automation driven by artificial intelligence. 2025-01-01T00:00:00Z http://yetl.yabesh.ir/yetl1/handle/yetl/4310972 An Large Language Model-Augmented Method to Assist Novice Designers in Divergent Thinking Lyu, Ke; You, Jiaxiang; Chen, Liuqing; Sun, Lingyun Divergent thinking is crucial for novice designers’ creativity and innovative problem-solving skills. Due to their limited knowledge base, novice designers struggle to master divergent thinking methods and connect personal knowledge with design experience. This study introduces a novel method that uses large language models (LLMs) to assist novice designers in expanding their divergent thinking. This method incorporates a two-layer hierarchical structure that links each of the four perspectives—memory, operation, scene, and property—to specific dimensions, facilitating the systematic generation of new divergent ideas. By combining LLMs’ vast knowledge with these divergent thinking frameworks, this method not only provides structured guidance but also fosters the gradual enhancement of divergence through interactive questioning and content generation. A series of tests, including the alternative uses test (AUT) and conceptual design tasks, were conducted to evaluate the impact of this method. Results show that LLMs may significantly improve the fluency, flexibility, and originality of divergent thinking outcomes. This study explores the potential for human–computer collaboration to support divergent thinking, opening avenues for future research and practical applications in design education and practice. 2025-01-01T00:00:00Z http://yetl.yabesh.ir/yetl1/handle/yetl/4310963 An Efficient Conflict Detection and Resolution Scheme for Geometric Constraints Using a Pruning and Backtracking Strategy Mu, Anyu; Liu, Zhenyu; Duan, Guifang; Tan, Jianrong Identifying and eliminating conflicts and redundancies between geometric constraints is crucial for effective constraint resolving in engineering design. This research proposes a graph-based conflict detection and resolution scheme for geometric constraint systems with both equality and inequality constraints based on numerical methods using a pruning and backtracking strategy. Initially, the minimum subset of conflicting constraints is detected by traversing all connected subgraphs of the original constraint graph in a pruning manner. The traversal process is encoded in a directed acyclic graph (DAG). The solvability of each constraint subgraph is determined by solving its equivalent algebraic system using variants of the Levenberg–Marquardt (LM) algorithm (Ma, 2008, “A Globally Convergent Levenberg–Marquardt Method for the Least l2-Norm Solution of Nonlinear Inequalities,” Appl. Math. Comput., 206(1), pp. 133–140; Amini et al., 2018, “An Efficient Levenberg–Marquardt Method With a New LM Parameter for Systems of Nonlinear Equations,” Optimization, 67(5), pp. 637–650) and verifying the solution. Inconsistencies between conflicting constraints are eliminated by modifying or discarding constraints recommended by a set of criteria. Finally, the resolution is validated by backtracking ancestor subgraphs along the paths of the DAG. Experimental results demonstrate the effectiveness of the proposed framework in handling inconsistent overconstrainedness between geometric constraints in various parametric forms, including those arising from violations of geometric rules or theorems. 2025-01-01T00:00:00Z http://yetl.yabesh.ir/yetl1/handle/yetl/4310892 An Image Generator Enhanced Deep Operator Network for Predicting the Geometry Deformations in Contact Problems With Random Rough Surfaces Liu, Daxin; Guo, Xuxin; Liu, Zhenyu; Tan, Jianrong Optimizing the geometry deformation characteristics in contact problems with random rough surfaces is an important component of improving product performance, such as assembly accuracy, sealing percolation, contact thermal resistance, and electrical resistance. Traditionally, the deformation is computed by numerically solving the partial differential equations that govern the contact problems. In the optimization process, the deformations under a variety of random rough surfaces need to be solved. It is computationally intensive and necessitates a surrogate model to approximate the numerical solutions. This study employs non-uniform rational B-splines (NURBS) to represent the geometries involved in the contact problem and proposes treating the NURBS control points as image pixels, treating the deformations of these points as image pixel values. Furthermore, an image generator-enhanced deep operator network (IGE-DeepONet) that leverages an image generator as a trunk net is proposed to predict the deformations and a concatenation-based information fusion mechanism between the trunk net and branch net of the DeepONet was developed to improve the prediction accuracy. Based on the contact problem between a smooth elastomer cube and a rigid cuboid with a random rough surface, it was demonstrated that the proposed IGE-DeepONet has smaller test error and reduced training time compared to the standalone image generator and the traditional DeepONet which uses a fully connected neural network as trunk net. 2025-01-01T00:00:00Z