SociableSense: Exploring the Trade- offs of Adaptive Sampling and - - PowerPoint PPT Presentation
SociableSense: Exploring the Trade- offs of Adaptive Sampling and Computation Offloading for Social Sensing Kiran K. Rachuri, Cecilia Mascolo, Mirco Musolesi, Peter J. Rentfrow Mateusz Grabowski The images come from an original presentation:
SociableSense: Exploring the Trade-
Sampling and Computation Offloading for Social Sensing Kiran K. Rachuri, Cecilia Mascolo, Mirco Musolesi, Peter J. Rentfrow Mateusz Grabowski
The images come from an original presentation: http://www.sigmobile.org/mobicom/2011/slides/58-socialsense-slides.pptx
How sociable am I? How to measure it quantitatively? How to provide real-time feedback to users about their sociability? How does information about my sociability affect my performance at work?
Microphone - Speaker Recognition Bluetooth - Colocation Detection Accelerometer - Activity Recognition
Frequency of sensor sampling
Computation performing
SENSING SLEEPING SENSING SLEEPING SENSING SLEEPING SLEEPING SENSING SLEEPING SENSING SLEEPING
SUCCESS: catched an interesting event FAILURE: catched a missable data ACTION: sensing from a sensor
SUCCESS: P := P + a(1 - P) FAILURE: P := P - aP PROBABILITY OF SENSING
0 < a < 1
PROB ACTION EVENT RESULT PROB ADJUSTING 0.5 sleeping missable
sensing unmissable success 0.5 + 0.5 * (1 - 0.5) = 0.75 0.75 sensing missable failure 0.75 - 0.5 * 0.75 = 0.375 0.375 sleeping missable
sleeping unmissable
sensing unmissable success 0.375 + 0.5 * (1 - 0.375) = 0.6875 0.6875 sensing unmissable success 0.6875 + 0.5 * (1 - 0.6875) = 0,84375 ... ... ... ... ...
a = 0.5 SUCCESS: P := P + a(1 - P) FAILURE: P := P - aP
Where to perform the computation? Which factors to consider when deciding? How to measure the impact of these factors on the future?
Let's consider:
and assume that they are pre-calculated and available before-hand!
Let's divide a task into the subtasks: T = [t1, t2, t3, ..., tn] There are 2n combinations in which T can be computed.
Configuration Subtask 1 Subtask 2 C1 locally locally C2 locally remotely C3 remotely locally C4 remotely remotely
uC(i) = wenergyue(i) + wlatencyul(i) + wdata over networkud(i)
ue(i) =
emin - e(i) e(i) emin = min {e(i) | i = 1, 2, ..., 2n}
wenergy + wlatency + w data over network = 1
<rules> <condition="battery_left > 80 and data_sent < 50MB"> <weight metric="energy">33.3</weight> <weight metric="latency">33.3</weight> <weight metric="data">33.3</weight> </condition> <condition="battery_left < 20"> <weight metric="energy">60</weight> <weight metric="latency">20</weight> <weight metric="data">20</weight> </condition> <condition="data_sent > 50MB"> <weight metric="energy">20</weight> <weight metric="latency">20</weight> <weight metric="data">60</weight> </condition> <condition="default"> <weight metric="energy">33.3</weight> <weight metric="latency">33.3</weight> <weight metric="data">33.3</weight> </condition> </rules>
Changing weights in time
○ Sensor Sampling Benchmarks ○ Computation Distribution Benchmarks
(i.e., X, Y, Z coordinates)
identifiers)
recordings)
Nokia 6210 Navigator phone
Sensor Unmissable event Missable event Microphone audiable data silence Bluetooth change in the number of colocated users no change Accelerometer movement of a user user is stationary
Accuracy: the percentage of missed events Energy consumption: measured using the Nokia Energy Profiler Latency: the delay in detecting change of event sequence from missable to unmissable and vice versa
Always the lower the better.
selected "a" parameter: 0.5
Type Back-off function Advance function Linear k * x x / k Quadratic x2 sqrt(x) Exponential ex ln(x)
+ dynamic adaptation technique
x - sleep time
Lifetime of the phone: the total time until battery gets completely discharged from the fully charged state Average data sent over the network: the average number of bytes sent by the system over the 3G network to process a sensor task Average latency: the average time taken for processing a sensor task
○ converting the recorded audio sample to Perceptual Linear Predictive (PLP) coefficients file ○ comparing the extracted coefficients file with the speaker models of all users
Conf AR CD SI1 SI2 Con f AR CD SI1 SI2 C1 L L L L C9 R L L L C2 L L L R C10 R L L R C3 L L R L C11 R L R L C4 L L R R C12 R L R R C5 L R L L C13 R R L L C6 L R L R C14 R R L R C7 L R R L C15 R R R L C8 L R R R C16 R R R R
Weights set wenergy wlatency wdata over network S1 1 S2 1 S3 1 S4 0.33 0.33 0.33
Network constraint: pij is the proportion of time i spent with j Two dimensions:
B A C
1.0 1.0 0.5 0.5 A (0.5)2 + (0.5)2 = 0.5 B (1.0)2 + (1.0 * 0.5)2 = 1.25 C (1.0)2 + (1.0 * 0.5)2 = 1.25
B C A D E
0.4 0.4 0.4 0.5 0.5 0.2 0.2 0.2 1.0 C 0.69 A 0.83 B 0.83 D 1.13 E 1.32
showing users: sociability, strength of their relations, activity levels and alerts about the users in sociable locations