real-time sensor networks with applications in cyber-physical systems
DESCRIPTION
Real-Time Sensor Networks with Applications in Cyber-Physical Systems. Manimaran Govindarasu Dept. of Electrical and Computer Engineering Iowa State University Ames, IA 50011, USA [email protected] http://ecpe.ece.iastate.edu/gmani. TALK OUTLINE. System-level Energy Management - PowerPoint PPT PresentationTRANSCRIPT
Real-Time Sensor Networks with Applications in Cyber-Physical Systems
Manimaran Govindarasu
Dept. of Electrical and Computer EngineeringIowa State UniversityAmes, IA 50011, USA
http://ecpe.ece.iastate.edu/gmani
2
TALK OUTLINE
System-level Energy Management
End-to-End Energy Management
Cyber-Physical System applications in Smart Grid
Conclusions
3
Embedded System
Battery
Embedded Device
• Processor - computation
• Network Interface -communication
• Others:
• Memory
• I/O
Energy is the most important resourceIt needs to be managed efficiently
4
Sensor Net Applications
Traffic monitoring system
Wireless Industrial Networks
Border Security
• Sense• Encrypt• Decrypt• Aggregate• Communicate
Wireless Sensor Network (WSN) – Data Aggregation Tree Model
5
Root/sinkEn
d-to
-end
dea
dlin
ei
(compute(Ti), Communicate(Mi))
(sense, compute, Communicate)
Leaf Nodes
6
Wireless Network
WSN – Mesh network model
A
B
DC
E GF
Energy Management at the Computing Subsystem
(considering all the tasks)
Energy Management at the Communication Subsystem
(considering all the messages)
Energy Management at the System-level
(both messages & tasks)
Computation
Communication+
WSN Challenges
Energy Management
Schemes
Duty Cycling Data Driven
Sleep-WakeupMAC with
duty cyclingData Reduction
Energy efficientData Acquisition
Joint SchedulingTasks & Msgs
Online Adaptation
State-of-the-Art in Energy Management
9
System-level Energy Management
Energy Management inNetworked Real-Time Embedded Systems
Computing subsystemCommunication subsystem System-level(Comp. + Comm.)
Cross-Layer Energy-aware Task Scheduling
(DVS, DPM)
Energy-aware Message Scheduling
(DMS; Power Adaptation)
Energy-aware System-level Scheduling
(DVS + DMS)
• Single-hop• Multi-hop
Embedded Device Energy Model
CPU Network interface
CPU Energy consumption Transmission energy consumption
b – modulation leveld – source-destination distanceL – message lengthW – channel b/w in HertzT – transmission time
Vdd – supply voltagef – CPU frequencyCC – CPU cyclesT – execution time
Wb
LT
BER
N
bLdE
b
ts
02
6
12
Energy-Time Tradeoff (CPU)
11
Vdd
V ’dd = ½ Vdd
En
erg
yE
nerg
yTime
Time
E/4
2T
T
E
DMS (radio): Energy depends …
12
Depends on distance,
transmission time.
Linearly dependent on transmission
time
13
Energy Model: energy vs. delay tradeoff
CPU
energ
y
Computation time of a task with CC CPU cycles
Com
m.
energ
y
Transmission Delay of message with length L
Dynamic Voltage Scaling (DVS)
t
2t
t’
2t’
(V1,F1)
Lower (V2,F2); processor slows down
(b1)
lower (b2)low trans. rate
Varying processor voltage (v) & frequency (f)
Varying message modulation (b)
Communication energy management
Computation energy management Dynamic Modulation Scaling (DMS)
14
Energy-aware combined scheduling of tasks & messages – the problem
Wireless Network( single hop )
T1
T3 T2
T4
m1 m2
m3m4
Complex periodic
tasks
Problem Statement
Given: ‘n` such complex Periodic tasks
Goal: (1) Schedule Tasks & Messages (2) Assign task frequencies & msg mod. levels
Objective: Minimize total system energy consumption.
Constraints: Meet all the deadline, precedence & ready time constraints.
Deadline = period = D
15
Energy-aware combined scheduling of tasks & messages – the solution
1. Task Mapping
2. Schedule local tasks on the nodes
3. Schedule msgs on the network
Feasible schedule (Energy unaware)
4. Assign modulation levels to messages &frequency levels to tasks.
Final Energy Aware Schedule
T1
M1
T2
T4
M2
T5
T3
T7
T6
Feasible Schedule
Use the slack to assign modulation levels to messages &frequency levels to tasksWhile guaranteeing:(1)Deadline(2)Precedence(3)Ready-time constraintsThis is an NP-Hard Problem
P1
P1
P2
Ch
Scheduling – Static & Dynamic
Shared wireless network
P1
P4P3
P0
Online PhaseOffline Phase
Task and message
parameters
Offline energy-awareStatic
Scheduling algorithm
Statically createdscheduleEnergy-Aware
Static Sched. AlgoEnergy-Aware
Dynamic Sched. AlgoOther scheduling
problems
System-level energy-time tradeoff Analysis
17
System-level Energy vs. Delay Tradeoffs
A B
Message should reach B before a deadline, D.
Compute
Communicate
TA
MA
D
∆
t1 t20
??
Com
m.
energ
y
Transmission Delay
t1
t2
(e1,t1)
(e2,t2)
t3(e3,t3)
System-level energy-delay tradeoffs
18
1. Subsequent gains decrease
2. All slack should not go to messages
Gain based Static Scheduling (GSS)
Insert all messages and tasks into set Q
Is Q empty ?
Pick up the highest energy gain entity
Reduce its performance
mode by one level ?
Reduce its performance mode by one level
Remove ei from Q exit
Yes
No
Yes
No
Gain Based Algorithm: Example
10 9 8 7 6 5 4 3
M1M2
Message Movement Table
20
400
300
200
T1
T1
Task Movement Table
Can I move tothe next col. ?
Yes
T1
0
f = 400
M1
b = 10
M2
b = 10
T1
0
f = 300b = 7
M2
b = 7
M1
400 300
Complexity: (nt + nm)(ntkt+nmkm)
Dynamic Slack Utilization – Distributed Algo
21
SharedWirelessMedium
M7
T1 T2 T6
T3 T4
M8 M9M10
P1
P2
Channel
• Goal: Utilize dynamic slack performance scalingto further reduce energy consumption
• Conditions:(1) Correctness – deadlines & precedence constraints(2) Overhead – no additional messaging
Dynamic slack
T1 T5
P2P1
Dynamic Slack Utilization
22
Shared wireless network
P1
P4P3
P0
Online Phase
Rules
Rules
Rules
Rules
Rules
• Use dynamic slack locally.
• Do not change the Finish times of any task/msg.
23
Effect of Channel Bandwidth
2. At Low b/w,Comp-only consumes lesser energy
1. As b/w increases,All schemesconsume lesser energy
3. At high b/w,Comm-only consumes lesser energy
4. Throughout, GSS performs better than comp-only and comm-only
24
Related Work
Research Focus Basic Idea References
Processor Energy Management
DVS based Task Scheduling. DPM policies.
[Aydin et al., Pillai et al.]
CPU + Memory + I/O DVS based Task scheduling
[Shin et al.]
Communication energy management
Power Adaptation, DMS, sleep-wakeup
[Schurgers et al.]
CPU + Network interface
Node Level, DVS + sleep/wakeup
[Bren at al.]
Computation sys. + communication sys.
Network Level with (DMS + DVS)
[Anil et. al. – TPDS 2008]
25
TALK OUTLINE
System-level Energy Management Problem
End-to-End Energy Management Problem [1]
Cyber-Physical System applications in Smart Grid
Conclusions
[1] G. Sudha Anil Kumar, G. Manimaran, and Z. Wang, "End-to-end energy management in networked real-time embedded systems," IEEE Trans. on Parallel and Distributed Systems, Dec. 2008.
Data Aggregation Tree – End-to-end guarantees
26
Root/sink
End-
to-e
nd d
eadl
ine
Problem• Given:
• Aggregation tree• for each node (i) – Ti and Mi
• Modulations: [bmin,bmax]• CPU Freq: [fmin, fmax]
• Objective: • Minimize total energy consumption
• Constraints:• end-to-end deadline (D)• precedence constraints
Leaf Nodes
(sense, compute, Communicate)
Solution Approach
27
Obtain a feasible schedule
Assign message modulation levelsand task frequencies
ME
MC MD
TD
MB
f = 400
b = 10
b = 10
b = 10
370
318
Assign Task Frequencies and Message Modulation Levels
While guaranteeingPrecedence, ready time and end-to-end deadline constraints
Solutions space
End-to-End Energy Management Problem
Continuous Model(not realizable in practice)
Optimal Solution NP Hard
HeuristicsScheduling Algorithms (GSA & EGSA)
Discrete Model(realized in practice)
Optimal: MILP formulation
(worst-case: non-polynomial)
28
Performance Evaluation Algorithms/schemes compared
Optimal: MILP solved using ILOG CPLEX 10.100 Proposed: Gain based Algorithm (GSA) Proposed: Extended gain based Algorithm (EGSA) Baseline: comp-only (only tasks are scaled) Baseline: comm.-only (only messages are scaled)
Simulation Parameters Bandwidth Radius factor (source – destination distance) Computational Load (cycles per task)
29
Effect of Communication Radius
30
1. At low distance,Comp-only consumes lesser energy
2. Energy consumptions increase as we increase radius
2. Throughout, GSA & EGSA are close to MILP
Effect of Computation Load
31
1. At low Comp. Load, Comm-only consumes less energy
2. Energy consumptions increase as we increase comp. load
2. Throughout, GSA & EGSA are close to MILP
32
Summary of Results
Communication energy consumption is NOT always the dominant factor
Computation energy ~ communication energy consumption
At low message modulation levels Low bandwidth channels Short-distance communication High computation load In some cases, computation energy consumption >
communication energy consumption
System-level energy savings >> component-level savings
20-50% improvement for evaluated conditions
33
TALK OUTLINE
System-level Energy Management Problem
End-to-End Energy Management Problem
CPS Applications in Smart Grid
Conclusions
Applications:
Critical infrastructure monitoring Automated traffic control
Home Area Networks
Ubiquitous healthcare monitoring
Cyber Physical Systems (CPSs)Cyber Physical Systems (CPSs)
35
Smart Grid: A Cyber-Physical System
Source: http://cnslab.snu.ac.kr/twiki/bin/view/Main/Research
Wireless Network Design and Fault DiagnosisWireless Network Design and Fault DiagnosisWireless Network Design and Fault DiagnosisWireless Network Design and Fault Diagnosis
Design a network for real-time data delivery in presence of latency and bandwidth constraints and an associated
fault diagnosis
Malicious Activity Faulty
Structure
Faulty NetworkExtreme wind,
temperature, ice, landslide
Electrical disturbances
Cost Optimization
Latency Constraints
Bandwidth Constraints
Network Design
Mechanical State
Estimation
Dynamic Line Ratings
Incoming sensor data
Malicious Activity Faulty
Structure
Faulty NetworkExtreme wind,
temperature, ice, landslide
Electrical disturbances
Maintenance Decisions
Contr
ol De
cision
s
Cost Optimization
Latency Constraints
Bandwidth Constraints
Network Design
Mechanical State
Estimation
Dynamic Line Ratings
Data Preprocessor
Incoming sensor data
Recommendationsto the Operator
Fault Diagnosis
Inference
Sensitivity Analysis
Possible causes for observed
effects
Possible effects for observed
causes
Impact of parameter change on component reliability
Bayesian Network reliability assessment framework
Malicious Activity Faulty
Structure
Faulty NetworkExtreme wind,
temperature, ice, landslide
Electrical disturbances
Maintenance Decisions
Contr
ol De
cision
s
Cost Optimization
Latency Constraints
Bandwidth Constraints
Network Design
Wireless Network Design for Transmission line monitoringWireless Network Design for Transmission line monitoring
1 2 3 N
Control Center, CC
Substation (SS1)
Substation (SS2)
Control Center
Sub-station 1
Sub-station 2
Wireless Links
Cellular Network
Cellular Tower
NormalTransmission
Tower
Level 3
Level 2
Level 1
SCADA Links
Cellular Enabled Transmission
Tower
Given a directed graph G = (V, E) and a set of N flows,Find a feasible path for each flow such that the sum of the cost of all the paths is minimized while respecting the delay and bandwidth constraints of each flow.
Given:Evidence, E = (e1, ..., ek ), where ek is observed state of variable Xi
Find: Probability of variable Xj
being in a certain state x = P(Xj = x |
E)
Bayesian Network Bayesian Network Fault Fault DiagnosisDiagnosis
Cause I
Cause II
Effect
Sample BN modeling a tower Sample BN modeling a tower Fault Fault DiagnosisDiagnosis
40
TALK OUTLINE
System-level Energy Management Problem
End-to-End Energy Management Problem
Cyber-Physical System applications in Smart Grid
Conclusions
41
Conclusions
System Level Energy Management offers significant savings CPU time, Communication, Memory, I/0
Commn (radio) energy is not always the dominant factor Depends on: modulation level, Sender-Receiver distance, Bandwidth
End-to-end Energy management while meeting deadlines Dynamic slack generation and utilization are key to energy savings
Cyber-Physical System poses constraints for network design End-to-end Latency, Bandwidth constraints, legacy comm links
Fault diagnosis distinguishes true faults from false positives
Future Work
Communication of Energy Management Leveraging physical layer techniques (Dynamic Code Size Scaling) Network coding + Energy-aware scheduling
System-level Energy Management Exploit sensing redundancy (temporal and spatial) more savings Holistic Scheme: CPU + Commn + Memory + I/O Distributed algorithms
Embedded sensor network design and operation (CPS) Self-healing, Security, Fault diagnosis, Decision Algorithms Applications of wireless sensor networks are endless !
42
43
Thank You !
AcknowlegementsSudha Anil KumarBenazir Fateh