Centralized and decentralized policies for the containment of moving source in 2D diffusion processes using sensor/actuator network

• Published in: 2009 American Control Conference pp. 127 - 132
• Main Author: Demetriou, M.A.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2009
• Subjects: Actuators; Centralized control; Diffusion processes; Distributed control; Employment; Energy management; Monitoring; Output feedback; State estimation; Switches
• ISBN: 142444523X; 9781424445233
• ISSN: 0743-1619; 2378-5861
• DOI: 10.1109/ACC.2009.5160089

In this note we consider centralized and decentralized control policies for the detection and containment of a moving source in 2D diffusion-advection PDE, often describing environmental processes. Such a task is enabled by the employment of a network of sensing devices judiciously located within the 2D spatial domain. These devices are assumed to have actuating capabilities aimed at containing the moving source by minimizing its effects on the process concentration. The source-detecting ability of the sensor network is considerably enhanced when the sensing devices are equipped to measure spatial gradients as opposed to only process concentration. The proposed estimation scheme provides estimates of the process state and at the same time provides an estimate of the proximity of the moving source. An added feature of the supervision and monitoring scheme is a power management scheme whereby a subset of the available sensors within the network are kept active over a time interval while the remaining devices are kept dormant. The resulting hybrid infinite dimensional system switches both the actuator, deemed more suitable to contain the source over the duration of a given time interval, and its associated control signal. Additionally, it switches the set of active sensors that are used by the scheme. The control policy examines two different schemes in which both a centralized and a decentralized scheme are considered. In the centralized scheme, information on the status of the active sensors along with the estimate of the state process are transmitted to the supervisor to feed a dynamic output feedback control signal to the actuator closest to the moving source. In the decentralized scheme, a computationally efficient controller is implemented, whereby the outputs from the active sensors are independently fed to the collocated actuators via static output feedback. Simulation studies utilizing at each time 16% of the total sensors and having either a single actuator with a centralized scheme or 16% of the total actuators with a decentralized scheme and used to minimize the effects of the moving source, are presented.