Multiple model robustification of iterative learning and repetitive control laws including design from frequency response data

Conference proceedings article

ผู้เขียน/บรรณาธิการ

เบญจมาศ พนมรัตนรักษ์

กลุ่มสาขาการวิจัยเชิงกลยุทธ์

ไม่พบข้อมูลที่เกี่ยวข้อง

รายละเอียดสำหรับงานพิมพ์

รายชื่อผู้แต่ง: Panomruttanarug B., Longman R.W., Phan M.Q.

ปีที่เผยแพร่ (ค.ศ.): 2009

Volume number: 134

หน้าแรก: 2259

หน้าสุดท้าย: 2278

จำนวนหน้า: 20

ISBN: 9780877035541

นอก: 0065-3438

eISSN: 0065-3438

URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-79952232132&partnerID=40&md5=8a205cd3172e67b45a68ac094eb25fa5

ภาษา: English-Great Britain (EN-GB)

บทคัดย่อ

Repetitive control (RC) and iterative learning control (ILC) aim for zero tracking error of a control system in repeating situations. RC can eliminate the influence on fine pointing equipment of slight imbalance in reaction wheels, and ILC can aim for zero error in repeated scanning maneuvers. Asking for zero error can put severe requirements on the accuracy of one's model of the system. One wants to achieve zero error in the hardware governed by the real world model and this can be different than our mathematical model used for design. Hence, stability robustness is important, asking the ILC or RC to converge to zero error in spite of imperfect knowledge of the system. Several previous publications have studied the averaging of a cost function over a distribution of possible models in order to improve stability robustness. This paper extends these works to give a more complete overview of the benefits of the approach. Three main classes of ILC are considered., and optimization criteria for each are generated so that one can perform the needed averaging. An important design approach in RC bypasses the use of a model, and directly uses experimental frequency response data. Instead of using multiple experiments to improve the frequency response information, a cost function is averaged over the data sets. Numerical investigations demonstrate for each case for ILC or RC, very substantial improvement in convergence to zero tracking error in the presence of model error.

คำสำคัญ