Portable and easy-to-use operating systems for wireless sensor networks

Transcription

1 This work has been supported by European Social Fund, grant Nr. 2009/0138/ 1DP/ /09/IPIA/ VIAA/004 Support for Doctoral Studies at the University of Latvia Portable and easy-to-use operating systems for wireless sensor networks Girts Strazdins

2 Riga -> Ålesund = 1200km / 2.5 hours 2

3 Ģirts Strazdiņš University of Latvia BSc.comp in 2006, Msc.comp in 2008 Dr.Sc.comp in 2012 (defense in dec/jan) Teaching Wireless Sensor Networks for undergrads 2010 and 2011 Institute of Electronics and Computer Science (EDI) Researcher Mikrotik: RouterOS programmer

4 Wireless Sensor Networks (WSN) Ģirts Strazdiņš PhD apspriedes seminārs

5 WSN node limited resources Energy: batteries CPU: 1-20 MHz RAM: 256 B 10 KiB Program code: KiB External flash: 1 MiB 4 GiB Tmote Sky sensor node 5

6 WSN programming abstractions 6

7 WSN programming abstractions 7

8 Contents of my thesis 1. WSN deployment survey, OS design rule inference 2. Analysis of existing solutions 3. Development of WSN OS: MansOS 4. Analysis of WSN object-oriented programming 5. Development of object-oriented WSN OS: OOMOS 6. Evaluation of MansOS and OOMOS 8

9 WSN deployment survey 9

10 WSN deployment survey Years articles from ACM, IEEE, Elsevier, Springer WSN in deployment and pilot-project stages Different applications: environment and animal monitoring, urban and military applications 10

11 Survey conclusion #1 Adapted or custom platforms in 80% of cases 20 % 48 % Adapted Custom Off-the-shelf 32 % 11

12 Survey conclusion #2 17% of users still create custom OS or program from scratch 8 % 5 % 17 % 10 % 60 % TinyOS custom Contiki other unknown 12

13 Survey conclusions #3 33%: only one user-level task (max 6 tasks) 20%: time-sensitive user tasks 57%: multihop routing used 38%: time synchronization used 13

14 Existing WSN operating systems (OS) TinyOS de facto standard, hard to learn Contiki increasing popularity, heavyweight LiteOS Unix-like, platform-specific Mantis Unix-like, not maintained Non-WSN OS: FreeRTOS no communication support Arduino simple, rapid, not energy-efficient 14

15 MansOS operating system 15

16 Effective, portable, easy-to-use WSN OS Developed at Univ. of Latvia and EDI, since

17 MansOS architecture 8/ 21/ 12 10:52 AM Application code HIL Radio Alarms Int. flash LED SPI Ext. flash Platforms & architectures Hardware-indep. implementations Function binding I/O port binding Platform constants Chips MCU Radio Flash Sensors MSP430 CC1101 M25P80 SHTxx Atmega CC2420 AT25DF ISL

18 MansOS features [1] Abstraction layers for portability Configurable, excludable modules code size optimization Three scheduling options: event handling, preemptive and cooperative wide application range Network protocol stack flexible communication Remote management and reprogramming rapid deployment and debugging 1. A.Elsts, G.Strazdins, A.Vihrovs, L.Selavo. Design and Implementation of MansOS: a Wireless Sensor Network Operating System. LU Scientific papers, University of Latvia, Vol.787, pp79-105,

19 MansOS use cases Smart car in GCDC challenge [2] EdiMote prototyping platform [3] Orchard monitoring (raspberries) [4] 2. G. Strazdins, A. Gordjusins, G. Kanonirs, V. Kurmis, A. Mednis, R. Zviedris, and L. Selavo. Team Latvia GCDC 2011 Technical Paper, in The Grand Cooperative Driving Challenge (GCDC 2011), Automotive Week 2011, Helmond, Netherlands, May 15-16, R.Ruskuls, G.Strazdins, and L.Selavo. Accurate Sensor Node Energy Consumption Estimation Using EdiMote Prototyping Platform. The 3rd International Workshop on Networks of Cooperating Objects (CONET 12), A. Elsts, R. Balass, J. Judvaitis, R. Zviedris, G. Strazdins, A. Mednis, and L. Selavo. SADmote: A Robust and Cost-Effective Device for Environmental Monitoring. International Conference on Architecture of Computing Systems, (ARCS 12),

20 MansOS use HIALS How ya doin? Networked robots? This ship cleanin is killin me! Picture origin: 20

21 Object-oriented OS 21

22 WSN object-oriented programming (OOP) Popular, convenient approach Thesis: it can be used for WSN OS design 22

23 Object-oriented MansOS (OOMOS) Increase code reuse by inheritance Specify inter-object interface using abstract classes Create objects with minimal interdepencencies Use only common C++ programming Implemented on TelosB, Zolertia Z1 and Arduino Duemilanove platforms 23

24 OOMOS MCU class diagram << Interface >> IMCU MSP430_GPIO MSP430MCU AVRMCU 1 AVR_GPIO 1 AVR_ADC MSP430_BasicClock MSP430x1xx MSP430x2xx MSP430_BasicClockPlus ATMega328P 1 MSP430_TimerA3 AVR_Timer16Bit MSP430_TimerB7 2 AVR_Timer8Bit MSP430F1611 MSP430F AVR_USART MSP430_USART MSP430_ADC12 MSP430_USCI_A MSP430_USCI_B 24

25 Adapting a new MCU << Interface >> IMCU MSP430_GPIO MSP430MCU AVRMCU 1 AVR_GPIO 1 AVR_ADC MSP430_BasicClock MSP430x1xx MSP430x2xx MSP430_BasicClockPlus ATMega328P 1 MSP430_TimerA3 AVR_Timer16Bit MSP430_TimerB7 2 AVR_Timer8Bit MSP430F2617 MSP430F1611 MSP430F AVR_USART MSP430_USART MSP430_ADC12 MSP430_USCI_A MSP430_USCI_B MSP430_ADC10 25

26 Provided interfaces Used interfaces 26

27 Evaluation: performance 27

28 Hz Sensor sampling (ADC) Single channel Multi-channel MansOS OOMOS Contiki TinyOS Only 8% of deployments required higher sampling rate 28

29 Sent, KB/s Radio TX performance Packet send throughput MansOS OOMOS Contiki TinyOS Packet size, B Max TX rate in deployments: 1.6KB/s (16 Byte 100Hz) 29

30 PRR, % Radio RX performance MansOS, OOMOS, TinyOS close to 100% Relative packet reception rate Packet size, B MansOS OOMOS Contiki TinyOS Contiki cannot handle its fast transmissions 30

31 Packets/sec Radio RX performance Absolute packet reception rate MansOS OOMOS Contiki TinyOS Packet size, B 31

32 Evaluation: memory size 32

33 Evaluation: static RAM 5000 Test application static RAM size, bytes Blink ADC Radio TX Radio RX MansOS OOMOS Contiki TinyOS 33

34 Evaluation: binary code size Test application code size, bytes Blink ADC Radio TX Radio RX MansOS OOMOS Contiki TinyOS 34

35 OOMOS optimization Unused inerfaces and components removed 100 % 100 % 24% 90 % 90 % Program code size reduction 80 % 70 % 60 % Functions 80 % 70 % 60 % 50 % Humidity 50 % 40 % Storage 40 % 30 % LogStream 30 % 20 % LEDs 20 % 10 % I2C 10 % 0 % Flash image size FastADC Minimal app 0 % Total Static RAM 35

36 Evaluation: portability 36

37 Lines of code OOMOS and MansOS reusability %+ reusable code Reusable Platform-dependent OOMOS MansOS Contiki TinyOS 37

38 Zolertia Z1 and TelosB platforms Component TelosB Zolertia Z1 Microcontroller MSP430, Series 1 MSP430, Series 2 Radio TI CC2420 TI CC2420 Sensors SHT11 humidity Analog light sensor SHT1 humidity TMP102 temperature, ADXL345 accelerometer External memory M25P16, 16 MBit M25P80, 8 MBit * Differences shown in bold 38

39 Lines of code Zolertia Z1 platform adaptation Zolertia Z1 platform code, lines of code Z1-specific Reusable added Reused TelosB OOMOS MansOS Contiki TinyOS 39

40 Conclusions MansOS and OOMOS are WSN OSes: Easy to use Actually applicable Easily portable OOMOS code size optimizations advisable Thank you! Girts Strazdins 40