Table 2: The approximation performance of the proposed SOM-Like inverse kinematics modeling method with five different network sizes.
Data | Set Network size | Average Error(mm) | Maximum Error (mm) | Computational Time (Sec.) |
Training Data Set | 5×5×5 | 36.15258494 | 119.3132718 | 12.3940467 |
10×10×10 | 12.57448587 | 59.03169383 | 13.8339731 |
15×15×15 | 7.013908311 | 70.72144217 | 18.9198033 |
20×20×20 | 4.965376925 | 51.22875539 | 33.0356172 |
20×30×15 | 4.957974223 | 60.27973511 | 53.5397344 |
Testing Data Set | 5×5×5 | 30.56174454 | 121.1298817 | 12.3965267 |
10×10×10 | 10.04231373 | 40.75206502 | 12.7682083 |
15×15×15 | 5.993293332 | 46.21338893 | 17.3399506 |
20×20×20 | 5.024934603 | 48.1309779 | 36.6631104 |
20×30×15 | 4.91614231 | 49.37489347 | 58.9459323 |