Distributed Scalable Control in Wireless Sensor Networks challenge statement

eggplantcinnabarMobile - Wireless

Nov 21, 2013 (3 years and 8 months ago)

90 views

Distributed Scalable Control in Wireless Sensor Networks

challenge statement

Michael Lemmon

Dept. of Electrical Engineering

University of Notre Dame

lemmon@nd.edu



Unreliable communication is a major obstacle to the d
evelopment of distributed and
scalable feedback control application over ad hoc wireless networks. For feedback
control, unreliable communication means that feedback data between sensors and
actuators are either delayed or dropped in a random manner. Th
e probabilistic nature of
these link failures suggests that ad hoc networks may be unsuitable for hard real
-
time
control systems. The challenge facing NSF’s ECS community is to determine whether or
not this is indeed the truth. In particular, the follo
wing questions immediately come to
mind. What fundamental limitation does an ad hoc wireless network impose on the
control system’s feedback channel? Are current formulations of control problems
sufficiently sensitive to the realities of wireless communic
ation in large
-
scale ad hoc
networks? How does one implement multiple feedback control systems on the same ad
hoc network? How do we guarantee control system performance in the face of network
unreliability? Do there exist novel performance indices that

can treat the optimization of
network QoS and control system performance in a unified manner? Can we use control
theory to identify novel quality
-
of
-
service (QoS) measures for ad hoc wireless networks
that can provide guidance for the development of new
communication protocols enabling
real
-
time feedback control over wireless networks?


Another important issue is the ability of ad hoc wireless networks to support real
-
time
supervision (as opposed to real
-
time control). Real
-
time supervision refers to d
iscrete
-
event logical supervisors that limit the freedom of intelligent systems (such as mobile
robots) so these systems do not damage themselves (or others) in executing a specified
task. NSF’s ECS community should develop novel supervisory control schem
es that are
distributed, scalable, self
-
stabilizing, and sensitive to real
-
time (as opposed to logical
-
time). In particular, the following questions come to mind. Is it possible to identify a set
of “canonical” supervisory actions that compose in a modu
lar manner, thereby avoiding a
need for verification of the entire network? What is the best way of developing
distributed self
-
stabilizing supervisors that are robust to missed messages over the
network? To what extent can real
-
time information be incl
uded in the analysis of such
supervisory control schemes? Is it possible to use these methods to develop novel
approaches for solving traditional problems in networked systems such as media access,
message routing, congestion control, and task scheduling?


There exists a need for testbeds used in decentralized control and identification. What
are the most appropriate testbeds for existing sensor network technologies? How can we
securely allow multiple researchers to use these testbeds?