Page 42 - ASME DSCC 2015 Program

Technical Program

Compensating flexibility in Servo Systems using Iterative learning Control of Multiagent formations: A Multiplex Information networks-
Control Based Approach
Contributed short paper. DSCC2015-9659 Contributed regular paper. DSCC2015-9644
Cong Wang, New Jersey Institute of Technology, Newark, NJ, United States, Dung Tran, Tansel Yucelen, Missouri University of Science and Technology,
Zining Wang, Tsinghua University, Beijing, China, Cheng Peng, Yu Zhao, Rolla, MO, United States
Masayoshi Tomizuka, University of California, Berkeley, CA, United States
This paper makes the first attempt to show how information exchange rules
Industrial servo systems usually have high gear ratio reducers, which intro- represented by a network having multiple layers (multiplex information
duce flexibility and transmission errors. The consequential vibrations and networks) can be designed for enabling spatially evolving multiagent for-
compliance bring difficulties to many demanding applications. For applica- mations. Toward this goal, we consider the invariant formation problem and
tions in which the servos move repetitively, iterative learning control (ILC) is introduce a distributed control architecture that allows capable agents to
a powerful tool to improve performance. An intuitive implementation of ILC spatially alter the resulting formation’s density and orientation without requir-
for flexibility compensation involves combining a torque ILC with a motor ing global information exchange ability. Efficacy of the proposed approach is
reference ILC. This paper explains why such a direct combination does not illustrated on numerical examples.
work well. A systematic synthesis method is introduced to address the cou-
Control of networked Multiagent Systems With uncertain Graph
pling between torque learning and reference learning. In addition, a robust
Topologies
control method is proposed to address system uncertainties. The proposed
Contributed regular paper. DSCC2015-9649
method is demonstrated using a servo control example.
Tansel Yucelen, J. Daniel Peterson, Missouri University of Science and
ConTRIBuTED SESSIon Technology, Rolla, MO, United States, Kevin Moore, Colorado School of
1-14-1 TA2 Multiagent network Systems 1 Mines, Golden, CO, United States
George Bellows B 10:00am–12:00pm
Multiagent systems consist of agents that locally exchange information
through a physical network subject to a graph topology. Current control
Session Chair: Hossein Rastgoftar, Drexel University methods for networked multiagent systems assume the knowledge of graph
Session Co-Chair: Tansel Yucelen, Missouri University of Science and topologies in order to design distributed control laws for achieving desired
Technology global system behaviors. However, this assumption may not be valid for
situations where graph topologies are subject to uncertainties either due
Detecting Behavioral Anomaly in Social networks using Symbolic
to changes in the physical network or the presence of modeling errors
Dynamic filtering
Contributed regular paper. DSCC2015-9643 especially for multiagent systems involving a large number of interacting
agents. Motivating from this standpoint, this paper studies distributed control
farshad Salimi naneh Karan, Subhadeep Chakraborty, University of of networked multiagent systems with uncertain graph topologies. The
Tennessee, Knoxville, TN, United States proposed framework involves a controller architecture that has an ability to
This paper investigates the use of Symbolic Dynamic Filtering (SDF) adapt its feedback gains in response to system variations. Specifically, we
algorithms in detecting anomalous behavior trends in social networks. analytically show that the proposed controller drives the trajectories of a
Data is generated from an agent-based discrete choice model, which networked multiagent system subject to a graph topology with time-varying
relies on a Markov Decision Process framework for stochastic simulation uncertainties to a close neighborhood of the trajectories of a given refer-
of decision-making in a social setting, where choices and decisions by ence model having a desired graph topology. As a special case, we also
individuals are influenced by social interactions. We show that such show that a networked multiagent system subject to a graph topology with
collective imitative behavior leads to rapid unstable fluctuations in the constant uncertainties asymptotically converges to the trajectories of a given
society, the fluctuation statistics being a weak function of the number of reference model. Although the main result of this paper is presented in the
extremist nodes present in the network as well as the prevailing political context of average consensus problem, the proposed frame- work can be
climate. In this paper, using a time-trace of global opinions in the said used for many other problems related to networked multiagent systems with
society, we investigate the effectiveness of SDF in estimating the number uncertain graph topologies.
of extremist nodes in a network, and studying the role of unpopular consensus of conspecific agents via collaborative and antagonistic
government policies as an enabler of political instability. Interactions
Spread of influence and ‘recruiting’ by extremist groups through social
Contributed regular paper. DSCC2015-9655
networks has become an important political issue in recent years. This study
Subhradeep Roy, nicole Abaid, Virginia Tech, Blacksburg, VA, United
is a step in the direction of building tools to preempt and intervene such
States
efforts.
The vast majority of the existing agent-based models of consensus consider
the interactions among the agents to be collaborative. In the present work,
we define superimposed stochastically-switching network topologies which
capture collaborative and antagonistic interactions among the agents. We
consider a general class of agents, so-called conspecifics, which encom-
passes a wide range of protocols explored in the literature, ranging from
42 Erdos-Renyi random networks to numerosity-constrained networks. We find
closed form expressions for necessary and sufficient conditions for consen-