Optimization Methods for Human–Machine Interaction Performance in the Design Phase

Authors:

Abstract
Digital twin models can comprehensively display the real-time operational status of industrial equipment and production lines through human-machine interfaces. However, in complex interaction scenarios, issues such as data lag and interface stuttering may arise in the interaction interface. These instances of virtual-physical asynchrony can lead to decision-making errors among operators, thereby limiting the effectiveness of the digital twin system in engineering applications. To address this issue, this paper proposes a Graphics Processing Unit (GPU) performance prediction model that integrates 3DMark scores. Concurrently, a multi-objective collaborative optimization strategy is established, with interaction accuracy and interaction real-time performance as the optimization targets. First, this method combines boundary GPU benchmark data with 3DMark scores to enable efficient prediction of interactive real-time performance for in-boundary GPUs. Then, the Cobb-Douglas function is introduced to establish a quantitative definition of human-machine interaction precision. Finally, the Non-dominated Sorting Genetic Algorithm II (NSGA-II) was employed, achieving synergistic optimization of accuracy and real-time performance, yielding a Pareto-optimal balanced solution. The method proposed in this paper is applied during the hardware selection and performance parameter design phase. It provides multiple optimal combination solutions for the performance of interactive interfaces under various operating conditions. To validate the effectiveness of the proposed method, verification was conducted across two typical industrial scenarios: forging production and logistics warehousing. Results demonstrate that, in the forging scenario, this method improves interaction time and interaction accuracy by 14.03% and 21.35%, respectively. In the warehousing scenario, these improvements are 4.22% and 2.28%, respectively.
Funding Information
  • National Science and Technology Major Project (2024ZD0608100)

This publication has 21 references indexed in Scilit: