ESO Architectures in the Trajectory Tracking ADR Controller for a Mechanical System: A Comparison
Proper operation of the Active Disturbance Rejection (ADR) controller requires a precise determination of the so-called total disturbance affecting the considered dynamical system, usually estimated by the Extended State Observer (ESO). The observation qu
- PDF / 1,684,118 Bytes
- 13 Pages / 439.37 x 666.142 pts Page_size
- 48 Downloads / 170 Views
Abstract. Proper operation of the Active Disturbance Rejection (ADR) controller requires a precise determination of the so-called total disturbance affecting the considered dynamical system, usually estimated by the Extended State Observer (ESO). The observation quality of total disturbance has a significant impact on the control error values, making room for a potential improvement of control system performance using different structures of ESO. In this article, we provide a quantitative comparison between the Luenberger and Astolfi/Marconi (AM) observers designed for three different extended state representations and utilized in the trajectory tracking ADR controller designed for a mechanical system. Included results were obtained in the simple simulation case, followed by the experimental validation on the main axis of a telescope mount. Keywords: Active Disturbance Rejection Control (ADRC) · Extended State Observer (ESO) · Trajectory tracking · Mechanical system
1
Introduction
Research interest in the Active Disturbance Rejection (ADR) control method, initially introduced in [4], has been gradually growing in recent years resulting in its frequent use in the domains of industrial- [10,14,18] and mobile robotics [13]. Instead of relying on the precise model of the control object, the ADR method depends on a feedforward cancellation of the so-called total disturbance, usually estimated by the Extended State Observer (ESO) with a single additional state, see [3]. Since the control quality obtained with the ADR-based controller is highly dependent on the total disturbance estimation accuracy, we may improve control performance by choosing the structure of ESO that is more suitable to the expected type of disturbance. Instead of increasing the observer gains to estimate quickly varying disturbances, one can implement an ESO augmented This work was partially supported by grants No. 33/32/SIGR/0003 and No. 2014/15/B/ST7/00429. c Springer Nature Switzerland AG 2020 A. Bartoszewicz et al. (Eds.): PCC 2020, AISC 1196, pp. 1323–1335, 2020. https://doi.org/10.1007/978-3-030-50936-1_110
1324
K. L akomy et al.
by multiple states, see [11], or use a Resonant Extended State Observer (RESO) [10] in the presence of total disturbance including an oscillatory component. An ESO is most commonly implemented according to a high-gain Luenberger observer design method [7], what usually results in a strong amplification of the sensor measurement noise. To decrease the impact of measurement noise Astolfi and Marconi introduced a limited gain power observer for the systems in the canonical observability form [1], which was generalized to a wider class of objects in [17]. In this article, we would like to compare the results obtained for three aforementioned observer structures, corresponding to a conventional ESO augmented by a single and multiple states, and a specific extended state representation utilized in RESO. Chosen structures will be implemented using two different observer architectures, i.e., a high-gain Luenberger observer and
Data Loading...
