Abstract
To address the optimization problem of excitation trajectories in robot dynamics parameter identification, this paper proposes an optimized objective function weighted by the condition number and maximum information entropy, utilizing an improved PID-based Search algorithm for trajectory optimization. First, an optimization model is established, where the objective function, weighted by the condition number and maximum information entropy, is constructed with the coefficients of the quintic polynomial and finite Fourier series as optimization variables, and the workspace is set as the constraint condition. Next, to enhance the dynamic adjustment capability of the PID-based Search algorithm for both local and global searches, the Adam algorithm is integrated into the PID-based Search algorithm for adaptive PID parameter tuning. The performance of the improved algorithm is verified using CEC2019 test functions, and the improved PID-based Search algorithm is employed for optimizing the excitation trajectories. Furthermore, comparative simulation experiments are conducted with the condition number as the objective function and with the condition number and maximum information entropy as the weighted objective function, using the improved PID-based Search algorithm, Based on reinforcement learning PID-based Search algorithm (RL PSA), PID-based Search algorithm (PSA), genetic algorithm (GA), and pattern search (Patternsearch) for verification. The superiority of the weighted objective function and the improved PID-based Search algorithm is validated through these comparisons. Finally, comprehensive experiments are conducted to confirm the effectiveness of the proposed method. The results demonstrate that the proposed method can obtain excitation trajectories with strong noise resistance and good generalization capability, significantly improving the accuracy and robustness of the overall robot identification results.
Keywords
Get full access to this article
View all access options for this article.