Cyber-Physical Systems
- 1. Towards Energy Efficient and Robust
Cyber-Physical Systems
Sinem Coleri Ergen
Wireless Networks Laboratory,
Electrical and Electronics Engineering,
Koc University
- 5. Wireless Networked Control Systems
Benefits of wireless
Ease of installation and maintenance
Low complexity and cost
Large flexibility to accommodate modification and upgrade of
components
Backed up by several industrial organizations
International Society of Automation (ISA)
Highway Addressable Remote Transducer (HART)
Wireless Industrial Networking Alliance (WINA)
- 6. Trade-off for Communication and Control Systems
Wireless communication system
Non-zero packet error probability
Unreliability of wireless transmissions
Non-zero delay
Packet transmission and shared wireless medium
Sampling and quantization errors
Signals transmitted via packets
Limited battery resources
Control system
Stringent requirements on timing and reliability
Smaller packet error probability, delay and sampling
period
Better control system performance
More energy consumed in wireless communication
- 7. Outline
Optimization of communication system given
requirements of control system
Novel design of scheduling algorithms
Joint optimization of control and communication systems
Novel abstractions for control systems
- 8. Outline
Optimization of communication system given
requirements of control system
Novel design of scheduling algorithms
Joint optimization of control and communication systems
Novel abstractions for control systems
- 9. Novel Scheduling Algorithm Design
Packet generation period, transmission delay and
reliability requirements:
Network Control Systems
sensor data -> real-time control of mechanical parts
Fixed determinism better than bounded determinism in control systems
(Tl ,dl ,rl )
- 14. Medium Access Control Layer: System Model
(Tl ,dl ,rl )
T1 £ T2 £ ... £ TL
given for each link l
Choose subframe length as for uniform allocation
Assume is an integer: Allocate every subframes
Uniform distribution minimize max subframe active time
Ti /T1 = si
T1
si
º
EDF
Uniform
max active time=0.9ms
max active time=0.6ms
✓
- 15. Example Optimization Problem Formulation
Maximum active time of subframes
Periodic packet generation
Delay requirement
Energy requirement
Maximum allowed power by UWB regulations
Transmission time
Transmission rate of UWB for no
concurrent transmission case
- 16. Outline
Optimization of communication system given
requirements of control system
Novel design of scheduling algorithms
Joint optimization of control and communication systems
Novel abstractions for control systems
- 17. Abstractions of Control System
Maximum Allowable Transfer Interval (MATI): maximum allowed time
interval between subsequent state vector reports from the sensor
nodes to the controller
Maximum Allowable Delay (MAD): maximum allowed packet delay
for the transmission from the sensor node to the controller
MAD MATI
Hard real-time guarantee not possible for wireless
-> Packet error probability >0 at any point in time
- 18. Abstractions of Control System
Stochastic MATI: keep time interval between subsequent
state vector reports above MATI with a predefined
probability to guarantee the stability of control systems
Many control applications and standards already use it
Industrial automation
IEEE 802.15.4e
Air transportation systems
Cooperative vehicular safety
Never been used in the joint optimization of control and
communication systems
- 19. Example Optimization Problem Formulation
Total energy consumption
Schedulability constraint
Stochastic MATI
constraint
MAD constraint
Maximum transmit
power constraint
- 20. Publications
Y. Sadi, S. C. Ergen and P. Park, "Minimum Energy Data Transmission for
Wireless Networked Control Systems", IEEE Transactions on Wireless
Communications, vol. 13, no. 4, pp. 2163-2175, April 2014. [pdf | link]
Y. Sadi and S. C. Ergen, “Optimal Power Control, Rate Adaptation and
Scheduling for UWB-Based Intra-Vehicular Wireless Sensor Networks”, IEEE
Transactions on Vehicular Technology, vol. 62, no. 1, pp. 219-234, January 2013. [pdf
| link]
Y. Sadi and S. C. Ergen, "Energy and Delay Constrained Maximum Adaptive
Schedule for Wireless Networked Control Systems", submitted.
- 21. Projects at WNL
Intra-Vehicular Wireless Sensor Networks
Supported by Marie Curie Reintegration Grant
Energy Efficient Robust Communication Network Design for
Wireless Networked Control Systems
Supported by TUBITAK (The Scientific and Technological Research
Council of Turkey)
Energy Efficient Machine-to-Machine Communications
Supported by Turk Telekom
Cross-layer Epidemic Protocols for Inter-vehicular Communication
Networks
Supported by Turk Telekom
- 22. Thank You!
Sinem Coleri Ergen: sergen@ku.edu.tr
Personal webpage: http://home.ku.edu.tr/~sergen
Wireless Networks Laboratory: http://wnl.ku.edu.tr
Editor's Notes
- Requires expertise for both systems, and still unsolved problem
- Scheduling design necessitates understanding requirements of sensor nodes and network
- If we had chosen a smaller subframe length
than T1, say T1=2, this may have resulted in a more uniform
distribution than choosing T1 still satisfying the periodic data
generation and delay requirements of the sensors. However,
since a transmission cannot be done partially in different time
intervals, the shorter unallocated time duration at the end of the
subframes may not allow including new nodes, changing the
transmission time or allocating additional messages violating
the adaptivity requirement. The shorter subframe length may
even avoid generating feasible schedules if the length of the
time slots is too large to fit in one subframe. Choosing the
subframe length larger than T1 on the other hand does not
bring any advantage and result in less uniform distribution.
- If we had chosen a smaller subframe length
than T1, say T1=2, this may have resulted in a more uniform
distribution than choosing T1 still satisfying the periodic data
generation and delay requirements of the sensors. However,
since a transmission cannot be done partially in different time
intervals, the shorter unallocated time duration at the end of the
subframes may not allow including new nodes, changing the
transmission time or allocating additional messages violating
the adaptivity requirement. The shorter subframe length may
even avoid generating feasible schedules if the length of the
time slots is too large to fit in one subframe. Choosing the
subframe length larger than T1 on the other hand does not
bring any advantage and result in less uniform distribution.