Papers »

Sparse estimation performance for wireless structural control

Structural control offers an attractive option for seismic protection by adding a supplemental device to introduce damping or alter the stiffness of the structure. Feedback control systems, which use measured responses from the structure to alter the supplemental device in real time, can adapt to changes in the structure and loading. Typically, a complete set of measurements or states are not available, so the structural system states are estimated for feedback control. With centralized, wired control systems estimation performance and error can be determined and limited with proper assumptions of measurement and modeling error. However, with wireless feedback systems, packet loss can lead to estimator and controller instability. Therefore, the estimator design and measurement feedback must account for packet loss to limit instability in a wireless control system. Due to their onboard sensing, communication, and computational abilities, wireless smart sensor nodes have become an attractive option to replace traditional tethered networks for structural control. For wireless sensor networks to be used in large-scale control systems, these challenges with wireless estimation must be understood.

The wireless benchmark control problem offers a unique test-bed to evaluate estimator performance and decentralized estimator design under packet loss. A sparse estimation design approach that weights the constant-gain estimator error covariance as well as measurement sparsity will be investigated on the benchmark problem. The sparse estimation feedback structure highlights the most important measurements for state estimation without requiring a trial and error approach. Additionally, a sparse form allows the control system to operate at a faster sampling rate, which can help overcome packet loss. Finally, an analytical solution to the bounds on steady-state error covariance under packet loss, which could be used in design, will be compared with the simulation results. Ultimately, the closed-loop wireless estimation performance will be compared with a wired centralized system.

Author(s):

Lauren Linderman    
University of Minnesota
United States