Iterative Learning Tracking for Multisensor Systems: A Weighted Optimization Approach.

Abstract

Multisensor systems are widely applied to realize the comprehensive monitoring and control as they feature multiple individual sensors/outputs. In such systems, different sensors can receive different types of operation signals, such as pressure, temperature, and volume. The desired references for different sensors may conflict in that an input signal that can precisely track all references simultaneously does not exist yet. This gap has motivated us to consider the incompatible multiobjective tracking problem for multisensor systems with random process disturbances and measurement noises. Our primary approach is to solve the problem as a weighted optimization problem using iterative learning control (ILC). First, the best achievable trajectory based on multiple references, as well as the weighted optimal tracking index, is carefully defined and then the ILC algorithms with both fixed and decreasing steps are proposed to generate the input sequence. The output driven by the proposed algorithms has been strictly proven to converge to the best achievable trajectory in both the mean square and almost-sure senses. Extensions to a networked implementation, in which the networks between the sensors and the learning controller suffer random data dropouts, are also detailed. Illustrative simulations are provided to verify the theoretical results.