Constrained optimization based control of real time large-scale systems: airjet object movement system

The control of real time, nonlinear, large-scale systems - systems with large aggregations of sensors and actuators - is seldom explored in actual operating physical systems. In such many-element systems, control issues such as actuation allocation, fusion of sensor data, and system identification emerge as challenging problems for large-scale system control. In this work, constrained optimization is used to solve these problems as applied to the control of an object moving system with 1,152 actuators and 32,000 sensors with a 2 ms control loop time. Solutions for allocating actuation among large numbers of actuators using hierarchical constrained optimization and fusing the output of many sensors into a small number of final measurements under tight real time constraints have been developed. This paper demonstrate that hyper-redundant systems are capable of system self-identification, and that constrained optimization can effectively solve problems associated with control of many-element systems