Generalized iterative learning control with mixed system constraints: A gantry robot based verification

Abstract

Iterative learning control (ILC) aims at improving the tracking performance of repetitive tasks based on information learnt from past attempts (trials). Modern practical applications demand more flexibility than current frameworks can deliver, in both how the task is specified and how system constraints are applied. To provide these features, an ILC framework is formulated in this paper for a generalized design objective with mixed system constraints, which includes intermediate position and sub-interval tracking as special cases. This is the first framework to combine a generalized ILC task description with constraint handling for continuous time systems. The successive projection method is applied to yield a comprehensive ILC algorithm with attractive convergence properties and computationally efficient implementation. This algorithm is verified experimentally on a gantry robot test platform, whose results reveal its practical efficacy and robustness against model uncertainty.

Iterative Learning Control for Path-Following Tasks With Performance Optimization

“…As normal practice, the pre-planned trajectory is defined by a fixed transition time [0.4, 0.8, 1.2, 1.6] ⊤ and the moving speed of the end-effector along each line segments is considered as a constant value. The final case implemented is to solve optimization problem (31) offline using the nominal plant model (61), yielding the solution plan [0.50, 0.84, 1.17, 1.50] ⊤ , and then to use this as a fixed reference in the standard successive projection ILC algorithm of [30]. The corresponding results of classical ILC and successive projection ILC are plotted in the same figures.…”

“…With the assumptions made in the theorem, the path following design objective of the problem (34) is within the range of the ILC problem discussed in [30]. Therefore, the ILC algorithm proposed in that paper can be applied with minor modifications to achieve the design objective with desirable convergence properties by setting the appropriate parameters, which give rise to the ILC update law (41)-(42).…”

“…The optimization of such a performance index brings significant practical benefits, such as control effort reduction, machine damage avoidance and increased manufacturing efficiency. Note that the authors' previous work in [27]- [30] has considered optimization of the intermediate time instant within point-to-point tracking problems and the path following problem with fixed intermediate time instants. However, they have not formulated a general ILC framework to address optimal path following problems in a principled manner.…”

Multiple Estimation Models for Discrete-time Adaptive Iterative Learning Control

“…Such a notion led to the design of numerous laws that constructed the control input in iteration (k + 1) based directly on the control input and error signal in iteration k. This eventually developed into a contraction-mapping (CM), operator-theoretic framework for ILC [2]. Many popular ILC strategies were designed based on this framework [3][4][5][6][7][8], and it continues to be popular, with numerous applications in large-scale industrial manufacturing [9,10], chemical batch processes [11,12] and robotics [13][14][15]. This framework has also been extensively analyzed, with established convergence and robustness results [16][17][18][19].…”

Robust PD-Type Iterative Learning Control of Discrete Linear Repetitive Processes in the Finite Frequency Domain

“…It merely requires some knowledge of dynamic characteristics and online calculation. Therefore, this technology has been widely used in the robot, power electronics, and chemical industries [3][4][5][6].…”