Local open-loop manipulation of multi-agent systems

Dinuka Sahabandu

Networks, or systems with distinct but interacting parts, are common in nature as well as in the engineered world. Emergence in networks, i.e. the internal dynamical interactions by which network components coordinate and achieve global functions, has been extensively studied. However, there is growing need to understand the impacts of external local actuation on network processes: for example, to understand the impacts of cyber- attacks on cyber-physical systems, study use dependence of neural processes, or design wide-area controls for networks. This thesis considers the manipulation of network dynamics from local actuation, using both theoretical and experimental approaches. First, manipulation of a canonical network-synchronization model known as the double-integrator network is studied analytically. For this model, conditions are given on the networks graph such that manipulation from a single location is possible. Further, the energy required to manipulate the network is characterized. It is shown that the manipulation energy closely depends on the spectrum of the graphs Laplacian matrix, the location of the actuation, and the goal state. In particular, manipulation along the consensus (synchronization) manifold is found to require little energy given sufficient time. Design of the double-integrator network model to make manipulation difficult (i.e., increase security) is also achieved. Second, an experimental study of the manipulation of neural circuits from local stimulation is pursued. This study is motivated by the need to understand the mechanisms underlying sleep regulation, which has been found to be an emergent network process. Here, small neural circuits grown in vitro are stimulated, and the evoked responses from these stimulations are analyzed. In particular, the impact on the evoked response of genetic and biochemical modifications of the neural circuits is determined. These modifications are found to alter evoked response shapes, which reflects altered functional states corresponding to sleep and waking.

Local open-loop manipulation of multi-agent systems

Files and links (1)

Abstract

Metrics

Details