Steering MAV Swarms

1
Steering MAV Swarms

• Demonstration
• Oleg Sokolsky
• Diana Gordon, William Spears, Insup Lee
• Patrick White, Tugkan Batu
• April 27, 1999

2
Monitoring of MAV pattern formation

• Demo is implemented as a concurrent Java
  application
• An alternative implementation of MaC framework
• the MaC objects are a part of the same
  application
• allows us to explore advanced features
• steering
• program checking
• The structure of MaC is followed closely
• Instrumentation is done by hand at source code
  level

3
Application Structure
4
Two avenues of monitoring

• Pattern formation
• monitored entity distance to a neighbor
• primitive event MAValert
  0.25R ? distance ? 0.75R
• monitor alarm pattern alarm
• trigonometric calculations
• monitored entities calls to functions sin and
  cos
• primitive events IoM(sin), IoM(cos)
• monitor alarm radiation alarm

5
Trigonometry checker

• Implemented at Cornell (Rubinfeld, White, Batu)
• Self-testing
• does the given code compute the required
  function?
• used by the monitor to check its own
  trigonometric computation
• Checking
• given numbers x and y, and a function f, is
  yf(x)?
• used by the monitor to check trigonometric
  computation in MAVs

6
Steering (repair)

• Monitor cannot address individual MAVs. All
  commands are broadcast by means of the Air class.
• Two types of steering are used
• after a pattern alarm, repulsion between close
  MAVs is suspended. MAVs are drawn together,
  re-starting the formation process.
• after a radiation alarm, computation of
  trigonometric functions is switched to a
  different mode.

7
Monitoring script (PEDL)
MonScr MAVpattern export event MAValert,
startPgm, checkSin, checkCos monobj int
Air.DIST monmeth void EmulateMAV.main(String)
monobj double MAV.run().dist monmeth double
MAV.myMath.sin(double) monmeth double
MAV.myMath.cos(double) event checkSin IoM(
MAV.myMath.sin(double) ) event checkCos IoM(
MAV.myMath.cos(double) ) event startPgm
startM( EmulateMAV.main(String) ) event
MAValert start( MAV.run().dist gt 0.25
Air.DIST MAV.run().dist lt 0.75
Air.DIST ) End
8
Monitored requirements (MEDL)
ReqSpec HexPattern import event MAValert,
startPgm, checkSin, checkCos var long
currInterval var int count0, count1, count2
var int prevAvg, currAvg event startPeriod
start(time(MAValert) - currInterval gt 10000)
property trigOK sinChecker.correct(value(checkS
in,1), value(checkSin,2))
cosChecker.correct(value(checkCos,1),
value(checkCos,2)) property NoPattern
(currAverage lt prevAverage1.15 100)
(prevAvg -1) startPgm -gt currInterval
time(startPgm) count0 0 prevAvg
-1 currAvg -1 startPeriod -gt
currInterval currInterval 10000 prevAvg
currAvg currAvg (curr0curr1curr2)/3 coun
t2 count1 count1 count0 count0 0
MAValert -gt count0 count0 1 End
9
Legend
Number of alerts goes down fast if everything is
normal
After a pattern alarm occurs, number of alerts
goes up due to correction
A radiation alarm occurs after a trigonometric
error is detected
A pattern alarm occurs after the number
of alerts grows too much
A bomb blast is shown in black
A radiation attack is shown in blue
10
Summary

• Demonstrate the integration of research results
  of the MURI participants
• steering of MAVs by means of artificial physics
• run-time monitoring of real-time systems with the
  MaC framework
• checking of trigonometric functions using program
  checking techniques