Multicore Applications at Data Fusion - Saab SDS

1
Multicore Applications at Data Fusion - Saab SDS
Dr. Mats Ekman
2
Saab Data Fusion Group

• A core team of about 18 engineers, including 6
  PhDs
• Active since 1984
• Air, Land, Naval, Civil domains
• Research Development
• Marketing/Sales support
• Technical tender support
• Analysis/Design
• Implementation
• Testing, customer training

xt1f(xt)wt
yt1h(xt)et
Parameter tuning Algorithm Redesign Alterations,
tests
Multi Sensor Tracker (MST)
3
Multicore ImplementationExample 1- a success
plots

• 2 step process
• get the positions
• calculate scalar products and compare with the
  plane
• Since objects are independent ?
• parallelization of the process
• TBB library (Intel Threading Building block) for
  C

tracks
sensor
4
Results

• Tested on a 4 cores ? local process 2.5 times
  faster.
• Delivered to customer - core 2.
• Drawback need to modify the code cannot use
  iterators. Some overhead using threading, cache
  misses?

Total process load
5
Example 2 a failure

• Association Process
• pre-processing transformation to polar
• coordinates and clustering
• Data association work on each cluster, since
  cluster are independent? parallelization

plots
tracks

• Technical problem
• Static variables several treads working
• on the same variables
• 2. Common resources ex. Id for tracks are
  obtained from a common track bank ?
• several treads in trying to access the bank ?
• lock (mute, sync)
• Solution restructure the code

sensor
Id bank
void set Void put
6
Ongoing and Future Multicore Applications at
Saab CoderMP cooperation

• Particle filtering
• Anomaly detection

7
Intro to particle filtering
A target here and now
prediction updating prediction updating
8
Probability densities
A target here and now
9
Filtering principles
Exactly Impractical
Ellipses/gaussian distributions Kalman
filtering
Particle filters
10
Particle filters
11
Comparison (1)
Standard Kalman
Constrained Kalman
Particle filter
12
Comparison (2)Particle filters - superior at
severe nonlinearities
Standard Kalman
Constrained Kalman
Particle filter
13
Parallelization of PFs
14
Videos

• Real Data from Enköping
• Acoustic Sensors
• No road constraints
• Simulated Data
• Acoustic Sensors
• Comparison between different road constrained
  filters
• Mix of real data from Gotland and simulated data
• Radar, acoustic and seismic sensors
• Road constraints
• Simulated Data
• Terrain constraints
• Comparinson with only road constraints

15
Anomaly detection complement to Rule Based
Situation Assessment

• Identify targets that do not behave like the
  majority
• Here Vessels south of Sweden.
• Blue Training data
• Green Test data identified as normal
• Red Test data identified as abnormal