Investigation of the Use of Commercial Robotic Arms for Real-Time Hybrid Substructuring 111 ref ref Fig. 8Result of an RTHS experiment using the KUKA® KR16 robot as an actuator without ILC (Trial 1) and after 25 ILC trials.The results are compared with the reference solution and with the results of an experiment using the Stewart platform as an actuator. 0 1 2 3 4 −4 −2 0 2 4 ·10−4 t / s etrack / m Stewart platform: Trial 1 Stewart platform: Trial 25 KUKA KR16: Trial 1 KUKA KR16: Trial 25 (a) The tracking error etrack over time for the different experimental setups without ILC (Trial 1) andafter 25 ILC trials. 0 5 10 15 20 25 10−3 10−2 Trial j e RMS,rel track Stewart platform KUKAKR16 (b) The relative RMS tracking error e RMS,rel track over the trials for the different experimental setups. Fig. 9Comparison of the tracking performance of the RTHS setup with the KUKA® KR16 as actuator with an experiment using the Stewart platform as an actuator. the additional task-space P controller used in the KUKA setup impeded ILC performance. Therefore, removing this controller may also improve the experimental results. Finally, the general performance of the ILC in the UDP communication framework needs further verification. Although we have shown in the previous section that the ILC can cope with large amounts of delay, the implementation on the hardware introduces further uncertainties, e.g. the processes on the MircoLabBox and on the KR C4 are not actively synchronized. In addition, the reliability of the internal processing on the KR C4 is unclear, which could lead to varying delays that could affect the performance of the ILC. CONCLUSION In this paper, we presented a framework to perform robust RTHS experiments using commercial robotic arms with high fidelity by combining ILC with NPC. The ILC aims to improve experiment fidelity by using iterative learning over multiple trials (a) The tracking error etrack over time for the different experimental setups without ILC (Trial 1) and after 25ILC trials. (a) The measured interface displacement z0meas of theRTHS experiments compared to the reference solutionzref. (b) The measured interface displacement F0ad of theRTHS experiments compared to the reference solutionFref. Fig. 8Result of an RTHS experiment using the KUKA® KR16 robot as an actuator without ILC (Trial 1) and after 25 ILC trials.The results are compared with the reference solution and with the results of an experiment using the Stewart platform as an actuator. 0 3 4 −4 −2 0 2 4 ·10−4 etrack / m Stewart platform: Trial 1 Stewart platform: Trial 25 KUKA KR16: Trial 1 KUKA KR16: Trial 25 (a) The tracking error etrack setups without ILC (Trial ) and after ILC trials. 0 5 10 15 20 25 10−3 10−2 Trial j e RMS,rel track Stewart platform KUKAKR16 (b) The relative RMS tracking error e RMS,rel track over the trials for the different experimental setups. Fig. 9Comparison of the tracking performance of the RTHS setup with the KUKA® KR16 as actuator with an experiment using the Stewart platform as an actuator. the additional task-space P controller used in the KUKA setup impeded ILC performance. Therefore, removing this controller may also improve the experimental results. Finally, the general performance of the ILC in the UDP communication framework needs further verification. Although we have shown in the previous section that the ILC can cope with large amounts of delay, the implementation on the hardware introduces further uncertainties, e.g. the processes on the MircoLabBox and on the KR C4 are not actively synchronized. In addition, the reliability of the internal processing on the KR C4 is unclear, which could lead to varying delays that could affect the performance of the ILC. CONCLUSION In this paper, we presented a framework to perform robust RTHS experiments using commercial robotic arms with high fidelity by combining ILC with NPC. The ILC aims to improve experiment fidelity by using iterative learning over multiple trials (b) The relative RMS tracking error eRMS,rel track over the trials for the different experimental setups. Fig. 9 Comparison of the tracking performance of the RTHS setup with the KUKA® KR16 as actuator with an experiment using the Stewart platform as an actuator. Substructure, is held constant over three time steps, while all other inputs to the NPC can change during these time steps. It needs to be investigated if this affects the performance of the NPC. For the ILC, it should be investigated whether using a zero-phase lowpass filter for Qinstead of the constant could improve the convergence of the ILC. Note that such a zerophase lowpass filter was used in the simulations of the previous section as well as for the Stewart platform setup. In addition, later simulations suggested that the additional task-space P controller used in the KUKA setup impeded ILC performance. Therefore, removing this controller may also improve the experimental results. Finally, the general performance of the ILC in the UDP communication framework needs further verification. Although we have shown in the previous section that the ILC can cope with large amounts of delay, the implementation on the hardware introduces further uncertainties, e.g. the processes on the MircoLabBox and on the KR C4 are not actively synchronized. In addition, the reliability of the internal processing on the KR C4 is unclear, which could lead to varying delays that could affect the performance of the ILC. Conclusion In this paper, we presented a framework to perform robust RTHS experiments using commercial robotic arms with high fidelity by combining ILC with NPC. The ILC aims to improve experiment fidelity by using iterative learning over multiple trials to mitigate compatibility errors in interface displacements, while the NPC ensures experiment stability by inducing artificial damping in case of instabilities. Both controllers act as pure outer-loop controllers in the task-space of the robotic arm, which makes the approach independent of the robot used, since no integration within the low-level controllers is required. In addition, our approach does not require any system modeling, allowing for easy implementation in different types of test setups. In virtual RTHS experiments, we demonstrated the ability of the approach to significantly reduce interface synchronization errors to provide high-fidelity RTHS results. We also showed its robustness to different amounts of delay in the Transfer System and different actuator models. We successfully implemented an RTHS experiment using a KUKA® KR16 as an actuator and integrated our control framework. The ability of the NPC to provide stable RTHS experiments could be verified, because without the use of the NPC, the designed experiment suffered from strongly growing unwanted oscillations. The performance of the ILC, on the other hand, has not yet been experimentally demonstrated. However, the presented experimental results are preliminary and further investigations are planned.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==