Bandwidth Allocation in Sense-and
Download
Report
Transcript Bandwidth Allocation in Sense-and
Bandwidth Allocation in
Sense-and-Respond
Systems
Vincenzo Liberatore
Research supported in part by NSF CCR-0329910, Department of Commerce
TOP 39-60-04003, NASA NNC04AA12A, and an OhioICE training grant.
Sense-And-Respond
Computing in the
physical world
Components
Sensors, actuators
Controllers
Networks
Sense-and-Respond
Enables
Industrial automation [BL04]
Distributed instrumentation [ACRKNL03]
Unmanned vehicles [LNB03]
Home robotics [NNL02]
Distributed virtual environments [LCCK05]
Power distribution [P05]
Building structure control [SLT05]
Merge cyber- and physical- worlds
Networked control and tele-epistemology [G01]
Sensor networks
Not necessarily wireless or energy constrained
One component of sense-actuator networks
Characteristics
Heterogeneous collection of networked sensors,
actuators, controllers
Power
Often plentiful, sometimes limited
Communication
Often wired, sometimes low-bandwidth wireless
Critical requirements:
Safety
Stability
Dependability
Robustness
QoS
Scalability
Adaptability
Information Flow
Flow
Sensor data
Remote controller
Control packets
Timely delivery
Stability
Safety
Performance
Outline
Outline
Introduction to Sense-and-Respond
Bandwidth Allocation
Future of Cyber-Physical Infrastructure
Warning
Most EE-oriented talk I could possibly give
Avoid redundancy with previous talks
Bandwidth Allocation
Bandwidth Allocation
Definition
Multiple sense-and-respond
flows
Contention for network
bandwidth
Desiderata
Stability and performance of
control systems
Must account for physics
Efficiency and fairness
Fully distributed,
asynchronous, and scalable
Dynamic and selfreconfigurable
Control and Networks
Control over Networks (Cover N)
NCSs, DCSs, SANETs, CPs, …
Control of Networks (Cof N)
Efficient BW allocation
Regulate the packet injection rate
“Cof N” scheme to better serve “Cover N”
Control of Networks
A bandwidth allocation scheme
Formulate the scheme as a Control
problem
Control systems regulate sending rate
based on congestion signal fed back from
the network
Sampling Rate and
Network Congestion
h=1/r
l1
l2
Problem Formulation
Define a utility fn U(r) that is
Monotonically increasing
Strictly concave
Defined for r ≥ rmin
Optimization formulation
max i Ui (ri )
s.t.
iS ( l )
ri Cl , l 1,..., L
and ri r min, i
Distributed Implementation
Two independent algorithms
End-systems (plants) algorithm
Router algorithm (later on)
Plant
Router
p
Controller
p
p
r ( pt ) 1 h U ' ( pt )
1
r max
r min
NCS-AQM Control Loop
Plant
Queue
1
r ( p) U ' ( p)
tf
q`=Σr(t) - C
Mt
o
d
b
q(t)
Queue
p(t)
f(q(t))
Controller
G(s)
Queue Controller G(s)
Proportional (P) Controller
GP(s) = kp
Proportional-Integral (PI) Controller
GPI(s) = kp + ki/s
q0
+
e
_
G(s)
u
q(s)
P(s)
Determination of kp and ki
Stability region in the ki–kp plane
Stabilizes the NCS-AQM closed-loop system for
delays less or equal d
Analysis of quasi-polynomials, f(s,es)
u(tj ) K ( R x(tj ))
Simulations & Results
50 Plants:
dx
ax(t ) bu (t )
dt
a bK a / r
U (r )
e
a
a
r min
ln bK a
bK a
[Branicky et al. 2002]
[Zhang et al. 2001]
Simulations & Results (cont.)
PI
¤
P
¤
Related Work
Congestion Control
Primarily addresses elastic flows
Active Queue Management (AQM)
Utility maximization and controllers often viewed as
alternative approaches
Multi-media congestion control
E.g., Equation-based
Smooth rate variation
No physically relevant utility
Time-scales
Approach to define time-varying utility functions
“C of N” missing
Outline
Outline
Introduction to Sense-and-Respond
Bandwidth Allocation
Future of Cyber-Physical Infrastructure
Warning
Most EE-oriented talk I could possibly give
Avoid redundancy with previous talks
Cyber-Physical Systems
Foundations and
technologies for rapid and
reliable development and
integration of computercentric physical and
engineered systems
“Globally virtual, locally
physical”
Major NSF initiative
planned
Needs and Directions
Needs and Directions
New Calculus
Merge time- and event-based systems
New Tools
E.g., co-simulation for co-design
New Networks methods
Bandwidth allocation, play-back buffers
New Education
Multi-disciplinary education
Telltale sign: New Metrics
Network-oriented metrics
Delay, jitter, loss rates, bandwidth
Impacts physics, but different from physics behavior
Control-Theoretical metrics
Overshoot, rise time, settling time, etc.
Hard to relate to network conditions
Multi-disciplinary metrics
E.g., plant tracking in terms of network bandwidth allocation
An E-Model for cyber-physical systems?
Example
PI
¤
P
¤
Acknowledgments
Students
Ahmad al-Hammouri
David Rosas
Zakaria Al-Qudah
Huthaifa Al-Omari
Nathan Wedge
Qingbo Cai
Prayas Arora
Colleagues
Michael S. Branicky
Wyatt S. Newman
Conclusions
Sense-and-Respond
Merge cyber-world and physical world
Critically depends on physical time
Bandwidth Allocation
Control of Networks to aid Control over Networks
Complete characterization of the stability region
Evaluation
Peak detection
Cyber-physical systems
http://home.case.edu/~vxl11/NetBots/