Overview of GME - Vanderbilt University

Download Report

Transcript Overview of GME - Vanderbilt University

Sensor Network-Based Countersniper System
Akos Ledeczi
Senior Research Scientist
Institute for Software Integrated Systems
Vanderbilt University
1
Copyright © 2004 Vanderbilt University
G. Simon, M. Maroti, A. Ledeczi, G. Balogh,
B. Kusy, A. Nadas, G. Pap, J. Sallai, K. Frampton
• Ad-hoc wireless network of cheap acoustic sensors is used
to accurately locate enemy shooters in urban terrain
• Performance:
– Average 3D accuracy: ~1 meter
– Latency: <2 seconds
– Multipath elimination
– Multiple simultaneous shot resolution
• Challenges:
– Severely resource constrained nodes
– Very limited communication bandwidth
– Significant multipath effects in urban environment
• Funded by DARPA through the IXO NEST program
2
Copyright © 2004 Vanderbilt University
Overview
Base Station
Shockwave
Muzzleblast
3
Supersonic
projectile
Copyright © 2004 Vanderbilt University
Technical Overview
• Detect TOA of acoustic shockwave and muzzle blast
• MICA2 mote
• Proprietary acoustic sensor board:
–3 acoustic channels (only a single channel is used in final system)
–High-speed AD converters
–FPGA for signal processing: shockwave and muzzle blast detection on board
•
•
•
•
Timestamp of shockwave and/or muzzle blast sent to mote
Motes send TOA data to base station
Base station fuses data, estimates shooter position and displays result
Middleware services:
–Time synchronization
–Message routing
–Remote control
4
Copyright © 2004 Vanderbilt University
Technical Approach
I2C
UART
Sensorboard
Time Sync
Muzzle Blast
&
Shockwave
Detector
SENSORBOARD
5
User
Interface
Sensorboard
Interface
Acoustic
Event
Encoder
Time
Sync
Sensor
Fusion
Sensor
Location
Sensorboard
Config/
Monitor
Data
Recorder
Message
Routing
Message
Center
Remote
Controller
Stack
Monitor
Remote
Control
Download
Manager
Plotter
Logger
MICA2 MOTE
BASE STATION
Copyright © 2004 Vanderbilt University
Software Architecture
Board
Clock
Mm1
τ1
ZC Coder
ZC Filter
Muzzle
blast
detector
τ2 = n/a
Mm2 = 0
τ3
ZC coder
Mm3
L2
L1
T1
6
I2C Interface
ADC
Shock
wave
detector
T2
T3
L3
time
Copyright © 2004 Vanderbilt University
Detection
• Sender-receiver multi-hop time synchronization
• Integrated leader election, global time is synchronized to the
local time of the leader
• End-to-end accuracy: average 1.6 μs per hop, maximum 6.1 μs
per hop (experiment included simulated root failure)
• Constant network load: 1 msg per 30 seconds per mote
• Start up time: network diameter times 60 seconds
• Uses the Time Stamping module
• Topology change tolerant: motes can move at speeds less
than 1 hop per 30 seconds.
• Available from the TinyOS CVS.
7
Copyright © 2004 Vanderbilt University
Flooding Time Synchronization Protocol (FTSP)
register
unregister
sent
received
aged
getRank
accept
• Flood Routing Engine:
–Ad-hoc routing
–Automatic aggregation
–Implicit acknowledgments
–Table/cache management
–Very low overhead
send
receive
User
User
Application(s)
Policy
Policy
Engine
• Flooding Policy:
OS / Radio stack
–Defines the meaning of “rank”
–Controls the flooding and
retransmission
• Application:
–Can change the packet on the way
–Can drop the packet on the way
Data packet:
– Fixed size length
– Must contain unique part
msg format:
8
app id
“rank”
3 bytes
packet 1
packet 2
packet n
Copyright © 2004 Vanderbilt University
Directed Flood-Routing Framework (DFRF)
• Combination of Time
Synchronization and
Message Routing
• No extra messages
• Stealth operation
• Uses the Time Stamping
module that has 1.4 μs
average precision per hop
• No clock skew estimation
• Precision depends on the
hop count of the route and
on the total routing time
• Plug-in replacement for the
Directed Flood Routing
Framework (DFRF)
9
Tevent
node1
node2
node3
Δt1
time
Δt2
time
Δt3
time
Δt1 + Δt2 + Δt3
time
root
Troot
Tevent = Troot - Δt1 - Δt2 - Δt3
Copyright © 2004 Vanderbilt University
RITS: Routing Integrated Time Synch
•
–
–
•
•
25%
Experiment #2: data is delayed by 5
seconds at each hop:
–
8 hours long (2 tests/min)
•
•
•
28.5 μs average error
107.6 μs average maximum error
265 μs peak maximum error
15%
10%
0%
0
2
3
5
6
8
10
11
13
synchronization error (microseconds)
maximum time synchronization error histogram
30%
1.5 hours long (2 tests/min)
4.4 μs average error
19.2 μs average maximum error
74 μs peak maximum error
20%
5%
simulates a shot event
triggered by a radio message in
experiment
•
•
•
10
30%
Root at the edge of the network
Experiment #1: normal routing:
–
•
average time synchronization error histogram
percentage
•
50 Mica2 motes
10 x 5 grid, neighbor to neighbor
comm is enforced in software
Five simulated shots separated by 10
msec
For each shot 13 motes send
simultaneous detection events to root
25%
percentage
•
•
20%
15%
10%
5%
0%
0
7
15
22
30
37
44
synchronization error (microseconds)
52
59
Copyright © 2004 Vanderbilt University
RITS Experimental Evaluation
Copyright © 2004 Vanderbilt University
Sensor Fusion
Shot #1 @ (x1,y1,T1)
t3
d3
Shot #2 @ (x2,y2,T2)
f(x,y)
t1
d1
?
Echo #1 @ (x3,y3,T1)
d4
t4
d2
t2
t3 – d3/v
t4 – d4/v
t2 – d2/v
11
t1 – d1/v
time
3
Shot time estimate T
sliding window
0
1
f(x,y) = [max number of ticks in window] = 3
• Advantages:
–Groups together consistent sensor readings
–Only uses correct detections for localization: high accuracy
–Enables multiple simultaneous shot resolution
• Search algorithm:
– Loop {
• Multiresolution search locates maximum
• If absolute time is close to a previously found peak, it is classified as an echo,
otherwise a shot
• Contributing sensor readings are removed
– } Continue
• Remarks:
–Size of sliding window is determined by the estimated detection error
due to, for example, sensor localization error
–Only uses muzzleblast at this point. Shockwave is utilized after
localization for trajectory estimation.
–Performance is remarkable: separates simultaneous shots, differentiates
between shooters in close proximity, can handle 10 shots per second or
more (bottleneck is network bandwidth)
12
Copyright © 2004 Vanderbilt University
Sensor fusion cont’d.
 Sep 2003: Baseline system
 Apr 2004: Multishot resolution
B1
NORTH
Church
 60 motes covered a 100x40m area
 Network diameter: ~7 hops
 Used blanks and Short Range Training
Ammunition (SRTA)
 Hundreds of shots fired from ~40 different
locations
 Single shooter, operating in semiautomatic
and burst mode in 2003
 Up to four shooters and up to 10 shots per
second in 2004
 M-16, M-4, no sniper rifle
 Variety of shooter locations (bell tower, inside
buildings/windows, behind mailbox, behind
car, …) chosen to absorb acoustic energy,
have limited line of sight on sensor networks
 Hand placed motes on surveyed points
(sensor localization accuracy: ~ 0.3m)
13
Copyright © 2004 Vanderbilt University
Experiments at McKenna MOUT site at Ft. Benning
Shooter Detection Error
40%
35%
percentage
30%
25%
2D
20%
3D
15%
10%
5%
0%
0
0.25 0.5
0.75
1
1.25 1.5
1.75
2
2.25 2.5 2.75
3
3.25 3.5 3.75
4
error (meter)
Based on 40 blank and SRTA shots from surveyed points
Average 2D error: 0.57m
Average 3D error: 0.98m
14
Copyright © 2004 Vanderbilt University
Results
Red circle:
 Shooter position
White dot:
 Sensor node
Small blue dot:
 Sensor Node that
detected current shot
Cyan circle:
 Sensor Node whose
data was used in
localization
Yellow Area:
 Uncertainty
15
Copyright © 2004 Vanderbilt University
2.5D Display, Single shot
Red circle:
 Shooter position
White dot:
 Sensor node
Small blue dot:
 Sensor Node that
detected current shot
Cyan circle:
 Sensor Node whose
data was used in
localization
Yellow Area:
 Uncertainty
16
Copyright © 2004 Vanderbilt University
2.5D Display, Multiple Shots
• New sensor fusion utilizing both muzzle blast and
shockwave:
– Increased range and accuracy
– Silenced weapons
• New sensor board:
– Low-power DSP
– More sophisticated detection: increased range
– Power saving modes
• Sensor self localization:
– < 0.5m 3D accuracy needed
• Scaling up:
– Hierarchical network architecture
– Distributed sensor fusion
http://www.isis.vanderbilt.edu/projects/nest
17
Copyright © 2004 Vanderbilt University
Future work