animal monitoring with unmanned aerial vehicle-aided

Animal Monitoring with Unmanned AerialVehicle-Aided Wireless Sensor Networks

Jun Xu, Gurkan Solmaz, Rouhollah Rahmatizadeh, Damla Turgut andLadislau Boloni

Department of Electrical Engineering and Computer ScienceUniversity of Central Florida - Orlando, FL

October 24, 2015

Jun Xu (UCF) LCN 2015 October 24, 2015 1 / 21

Outline

1 IntroductionMotivation & problem statement

2 Problem analysisClusteringNetwork modelingValue of information

3 Proposed path planning approachMarkov decision processPath planning process for UAV

4 Simulation studySimulation setup and metricsDemo displayPerformance results

5 Conclusion


Motivation

Animal monitoring have various goals:

Tracking their migration paths

Predict if specific endangered species exist

Goal of this application:

Providing reliable animal appearance information in large-scale areas

Do not using mounting devices & not affecting animal activities


Problem statement

How to find these animals?

Sensors can not directly send data to remote base station

How the sink (UAV) knows which sensors have the relevant information

How to use those sensed information?

Latency between animal appearance and information being gathered

How to quantify this information


Clustering

Real movement trajectories of 4 zebras in 3 days †Wildlife animals are more likely to having activities in a small area

†Yong Wang, Pei Zhang, Ting Liu, Chris Sadler, Margaret Martonosi, CRAWDAD dataset princetonzebranet (v. 02/14/2007),traceset: movement, downloaded from http://crawdad.org/princeton/zebranet/20070214/movement, doi:10.15783/C77C78,Feb 2007.


Clustering

Real movement trajectories of 4 zebras in 3 days †Wildlife animals are more likely to having activities in a small area



Network model

Cluster-heads are responsible forreceiving data from other sensorsand submitting data to the UAV

Single UAV communicates withcluster-heads


Value of information

The value of information (VoI)†

Sensed information has the high-est value when event occurs

Our goal is maximizing the VoIin the whole network

A: the initial value of the informationB: the decay speed of the VoI

†Turgut, Damla, and Ladislau Boloni. ”A pragmatic value-of-information approach for intruder tracking sensor networks.”Communications (ICC), 2012 IEEE International Conference on. IEEE, 2012.


Markov decision process model

5-tuple {S ,A,P,R, γ}:S is the set of states (grids) inthe network

A is the set of possible actionsthat UAV can do

P is the state transition probabil-ities

R is the instant reward when theUAV enters one gird

γ ∈ [0, 1) is the discount param-eter


Markov decision process model

Solved this MDPmodel by Q-learning

9 possible actions ofS4: 8 neighbors andstaying itself

Q(s, a) = R(s) +γmax

a′Q(s ′, a′)

Instant reward R(s),future potential re-ward Q(s ′, a′)

Possible actions of S4: {Northwest, North, Northeast, West, Stay, East,Southwest, South, Southeast}


Path planning flow chart

Exploitation: deter-ministic grid selectionby Q(s, a)value

Exploration: randomgrid selection

ε: random selectionprobability


Simulation setup

Movement traces of zebras:I ZebraNet project †I 5 zebras in June 2005 at a 10km × 10km area near Nanyuki, KenyaI 5682 GPS records in totalI GPS sampling time interval: 1 minute

Definition of sensing events:I If zebra switches grid, record the eventI If zebra always stays in one grid, record every ∆t time



Simulation setup

Simulator:I Java-based discrete time simulator

Performance metrics:I Value of informationI Average message delayI Number of zebras encountered

Approaches for comparison:I GreedyI Traveling salesman problemI Random


Simulation setup

Network size 10km × 10kmNumber of grids (states) 16Grid size 2500m × 2500mUnit experimental time (round) 10sUAV speed 100m/roundDecay speed of VoI (parameter B) 0.05Radius r for direct observation 200mInitial reward IR (σ, Ci , Idist , Iduration) 10.0 (10.0, 1.0, 1.0, 1.0)


Demo display


Value of information

0 500 1000 1500 2000 2500 30000

200

400

600

800

1000

1200

1400

Simulation time (min)

Val

ue o

f inf

orm

atio

n

MDPGreedyTSPRandom

MDP (Markov decision process)

Greedy (Greedy total number of pre-vious events)

TSP (Traveling salesman problem)

Random (Random selection from allgrids)


Average message delays

MDP Greedy TSP Random0

10

20

30

40

50

60

Ave

rage

del

ay (

min

)





TSP: 0 deviation because fixed route


Number of zebras encountered

MDP Greedy TSP Random0

50

100

150

200

250

Num

ber

of z

ebra

s en

coun

tere

d





Direct observation radius (r)


Performance stability

0 500 1000 1500 2000 2500 30000

500

1000

1500


Val

ue o

f inf

orm

atio

n

MDP−1MDP−2MDP−3MDP−4

Results from 4 time experiments

Same parameters


Exploration

0 500 1000 1500 2000 2500 30000

500

1000

1500


Val

ue o

f inf

orm

atio

n

ε = 0.2ε = 0.4ε = 0.6ε = 0.8

ε− Greedy policy

The probability of random grid selec-tion by UAV

Impact of exploration


Conclusion

We focused on the animal monitoring in large area

We proposed a MDP-based approach for UAV path planning

The evaluation indicated significant improvement compared to Greedy,TSP and Random

Future work:I Other species, other datasetI Multi-UAVs


animal monitoring with unmanned aerial vehicle-aided

Documents