Transcript UAV’s: Scenarios

UAV’s: Scenarios

Smart bombs
 no

need to return to base
Reconnaissance/surveillance
 types
of sensors used, mission specifics
(e.g. round trip, refuel & return)

Power source
 rechargeable
battery (by propulsion)
 standalone battery
 running off the generator directly

Application
 SAR

- resolution, speed of aircraft, type of aircraft, number of aircraft
Cooperating fleet
 homogeneous
vs. heterogeneous
Deployment Stage Scenario
Parachutes
Ejection

Ejection: Air carrier charges battery, downloads
software, then launches weapons group near targets at
high speed.

Drift: Radar and acoustic sensors locate potential
targets.

Ascent: Power boost propels the weapon upward to
increase target seek time.

End Game: Individual targets are selected and IR
sensor provides precision guidance.
Endurance 3-7 min
How to get description of
application?

System level description
 block
diagram level (UML)
 dataflow (UML annotation) / control flow (UML maybe)

Data sets
 sensory

UAV specific
 smart






bomb - stage scenarios:
GPS / target detection
inflight avoiding obstacles
on target
final steering towards target - propulsion control
communication requirements
Power source
 voltage,

traces / activations
power rating
Weight/volume constraints
Appropriate metrics

MIPS per watt
 higher
MPW=> can fly faster and survey larger areas
 keep algorithm same

Task per watt (or joule over time)
 peak
power, current limitations
 task completion time
 communication power

Computation resources required
 parts
cost, system cost
 weight
 maximum clock rate, memory footprint, secondary storage
 busses/wires count/weight
Baseline?

Existing working system in field
 extrapolated
missile system from Northrop Grumman
 Los Alamos - defining radar system on UAV
 Lincoln Lab RASSP-SAR benchmark

Tradeoffs
 Communication
vs. computation:
modular simulation environment for modeling communication vs.
computation issues
 Sensor power vs. computation or communication power
(clean the noise with computation rather than re-sense)