Esense: Communication through Energy Sensing. Kameswari Chebrolu, Ashutosh Dekhne Department of CSE, IIT Bombay

Transcription

1 Esense: Communication through Energy Sensing Kameswari Chebrolu, Ashutosh Dekhne Department of CSE, IIT Bombay

2 Esense: The Concept A variety of standards in the same spectrum b/g, , Bluetooth, Cordless Standards have very different physical layers Bits cannot be interpreted across devices following different standards Can interfere with each other Goal: Enable communication between devices with fundamentally different physical layers

3 Applications: Coordinated Coexistance WiFi AP Esense Comm: Inform WPAN of WiFi free periods Nearby WPAN

4 Applications: Energy Management WiFi AP Smart Phone 1 Esense Comm: Wake up Smart Phone 1 (not any other phone) Smart Phone 3 Smart Phone 2

5 Applications: Network Management Comm. Range Carrier-Sense Range AP1 Esense Comm: If you hear me (AP1), you are in my interference range AP2

6 Solution Approach Based on Energy Sensing (Esense) Three possible modulation techniques RSSI: Variable and environment dependent Inter energy burst gap: cannot control in CSMA Length of energy burst: sounds promising Amplitude Length of burst Inter Burst Gap Received Signal Strength Proof of Concept: Enable uni directional communication from an radio to an radio

7 An Example: Energy Management Esense Comm: Wake up Phone with ID4 (not any other phone) ID1 720 bytes WiFi AP Message Packet Length (bytes) Wake up ID1 200 Wake up ID2 450 Wake up ID3 600 Wake up ID ID2 ID3 ID4 ID5 ID6 ID7 ID8 ID9 Phones at Association obtain Esense message interpretation information

8 Challenges What possible energy burst lengths (Esense alphabet set) to use? Distinguish between Esense packets and Regular packets Accuracy of reciever hardware in detecting energy lengths g can support data rates upto 54Mbps Inter packet gap can be as small as 50/28us

9 Hardware Inaccuracy Packet not detected Packets are merged Predicted duration (3 clock ticks) smaller than actual Predicted duration (5 clock ticks) larger than actual 1 clock tick = 30.5us

10 Overall Approach Distingushing between Esense and Regular packets Leverage bi modal pkt size distributions in wifi networks Exclude packet sizes that occur with a frequency greater than a threshold % Allocate rest to Esense alphabet Alphabet size can be as large as 2288 ( ), provided mote can provide byte level resolution

11 Accuracy Characterization Actual channel Occupancy: 785us

12 Accuracy Characterization Actual channel Occupancy: 785us

13 Accuracy Characterization Actual channel Occupancy: 785us Conclusions: Inter Packet gap should be atleast 90us Packet duration error can be +/ 61us

14 Practical Alphabet Extraction multiple data rates; Inaccuracy of hardware Measurement based Alphabet extraction (for a given rate of operation): Use a real Wifi node and transmit packets Of size as dictated by the packet traces Interval as dicated by the CSMA MAC (backoff) Use hardware to measure run lengths (of clock ticks) of energy duration Extract Esense alphabet from the complement set of run lengths that exceed a given frequency threshold

15 Cafeteria Trace: 6Mbps

16 Practical Alphabet Extraction Bounding the Complement Set Maximum packet size (2304 bytes) sent at the lowest rate at which an node operates (g~6mbps, b~1mbps) Ensure margin tick seperation (between adjacent alphabets Use repetition to handle false positives and false negatives Esense packets always sent at the lowest rate permitted (g~6mbps, b~1mbps)

17 Experimental setup snapshots of 500 packets each of 5 different packet size traces (cafeteria, library, CSE, Conference) Back to back transmission (gap as dictated by CSMA MAC) Different data rates including random rates (to emulate auto rate adaptation) Threshold frequency 1%, margin: 4 ticks

18 Alphabet Size Trace 1Mbps 11Mbps 6Mbps 54Mbps All Rates CSE PSU (15) 118 (13) 107 (12) Library (15) 117 (14) 110 (12) Cafeteria (15) 118 (15) 110 (11) CSE Sford (14) 118 (14) 109 (12) OSDI (15) 119 (11) 110 (11) All traces (10) 117 (15) 100 (10) The numbers in brackets refer to g mode of operation i.e complement set bounded by 2034 bytes sent at 6Mbps

19 Validation 60 snapshots of non intersecting 500 packets corresponding to 2 different trace files 250 random alphabets inserted into a single trace file Sender Repetition Count: 1, 3, 5, 10 Receiver Detection Count: 1, 2, 3, 5 Random number of regular packets (max 5) between repetitions Note: Esense Packets are always sent at th

20 False Positives/Negatives B Compatible Mode G Compatible Mode 11 Mbps 18 Mbps R Adapt 18 Mbps R Adapt FN FP FN FP FN FP FN FP FN FP (1,1) (3,2) (5,3) (10,5) Cafeteria Trace

21 Discussion Applicability in a variety of contexts Other standards, bi direction communication Building vocabulary on top of base alphabet Lot of scope for future research Non bi modal packet size distribution Higher accuracy hardware, vocabulary, manipulate the size of regular packets Native support for Esense in future wireless standards