SLIDE 1
SensIT CSP Workshop Day 2 J anuar y 16, 2001 P alo Alto, CA - - PowerPoint PPT Presentation
SensIT CSP Workshop Day 2 J anuar y 16, 2001 P alo Alto, CA - - PowerPoint PPT Presentation
SensIT CSP Workshop Day 2 J anuar y 16, 2001 P alo Alto, CA Schedule 8:30 9:00 Br eakf ast 9:00 9:30 I ntr o Commons 9:30 10:30 Tr acker Design Aquar ium: # 5 (2-way tr af f ic) Commons: # 7 (per imeter violation)
SLIDE 2
SLIDE 3
Tracker Discussions
The Task: Develop a t op-level, end-t o-end solut ion f or t wo benchmar k pr oblems. I dent if y all t he pieces. Why:
- SenseI T t eams come f r om mult iple f ields, t ar get ing
dif f er ent aspect s of t he pr oblem:
- This is an oppor t unit y t o
–Get all of t he common t echniques on t he t able and discuss inf or mally –I dent if y any t echnology gaps –See how t r acker design, signal pr ocessing and classif icat ion appr oaches complement each ot her –Syner gies bet ween t echniques her e may make good demo candidat es Not e: t hese ar e j ust st ar t ing point s, f eel f r ee t o modif y scenar ios in int er est ing ways or t o f ix bugs.
SLIDE 4
Issues / Subtasks
- Quiescent St at e
- Tar get det ect ion / wakeup
- Clust er For mat ion
- Signal P
r ocessing – Dat a Associat ion – Sensor Fusion – Classif icat ion – Localizat ion – Compr ession
- Tr acking
– Tr ack gener at ion / dest r uct ion – Tr ack coalescence / split t ing – Dynamics – Reasoning about per sist ent ident it ies – Reasoning about Global P r oper t ies
- Exf ilt r at ion
- Quer y P
r ocessing
SLIDE 5
Scenario # 5: Two-way traffic
Task
- Track target positions
- Estimate target crossing time
Challenges
- Vehicles in close proximity,
need to use dynamics to keep identities separate
Task-Specific Benchmarks
- Accuracy of crossing time
estimate
- Nom. Crossing
Point
SLIDE 6
Scenario # 5 Assumptions
- Desir ed out put updat e r at e: 0.5 Hz
- Desir ed posit ion r esolut ion: 25’
- Net wor k behavior
– Lat ency: mean 0.1 secs per hop – P acket s may ar r ive out -of -or der – 10% packet loss – 19.2 kbps per clust er , each node can j oin up t o 2
- I nit ial Velocit ies: 25 f t / sec (each)
- Dynamics
– Max acceler at ion: +/ -50 f t / sec^2 (symmet r ic) – Not e t hat t anks can pivot 360 degr ees in place while at st andst ill
- Sensor spacing: 100’ bet ween all sensor s
- Vehicles: I dent ical, signat ur e aspect dependent
- I sot r opic pr opagat ion of sound, vibr at ions, no wind,
no inver sion layer – 1/ r ^2 at t enuat ion
- Tar get class known a pr ior i, no gear shif t s
SLIDE 7
Scenario # 7: Perimeter Violation Sensing
Task
- Alert on violation of perimeter
- Ignore activity outside of
perimeter (distractors)
- Identify violator type and track
location
Challenges
- Filter out distractor
- Respond quickly while
minimizing quiescent activity
Task-Specific Benchmarks
- Detection delay
- Power usage during periods
- f no violation
- Frequency of false positives
all vs. distractor/violator source amplitude ratio
1000’ 25’ arrival at t=0 arrival at t=8:00:00
SLIDE 8
Scenario # 7 Assumptions
- Desir ed det ect ion t ime: <
2 sec
- Dist r act or ar r ives, or bit s f or 8 hour s pr ior t o int r uder
- Dist r act or / I nt r uder sour ce amplit ude r at io: 40 dB
- Const ant velocit ies:
– I nt r uder 20 f t / sec – Dist r act or 40 f t / sec – I nt r uder can come f r om any dir ect ion
- Net wor k behavior :
– lat ency: mean 0.05 sec/ hop – P acket s may ar r ive out -of -or der – P acket Loss: 2% – 128 kbps/ clust er , each node can j oin up t o 4
- Sensor spacing: 100’ bet ween all sensor s
- I sot r opic pr opagat ion of sound, vibr at ions, no wind, no
inver sion layer – 1/ r ^2 at t enuat ion
- I nt r uder & Dist r act or classes unknown a pr ior i, no
gear shif t s, const ant speed
SLIDE 9
2001 Demo Discussions
The Task
- Develop candidate demos f or Mar ch/ Apr il and
November . I dentif y inter ests of var ious teams. Constr aints
- Mar ch: Field Exper iment using cur r ent nodes
- Apr il: Lab Demo, Simulation and/ or Data
exper iment using 29P alms or Mar ch 2001 dat a
- November : Field Exper iment