On-the-fly Route Planning for Mono-UAV Surveillance Missions M. Soulignac 1 F. Gaillard 1 C. Dinont 1 G. Marchalot 2 1 ISEN-Lille 2 THALES Airborne Systems UK PlanSIG'13 Edinburgh, 29 January 2014
Mono-UAV Surveillance Mission Planning Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
A Mono-UAV surveillance mission Detected target Non-detected target Identified target Targets UAV (unknown position and route) Identification range Maximal targets velocity Detection range known a priori 3
Mission Planning Expected output p D : % of detected targets p I : % of identified targets p D near 100% and p I near 100% Beginning of the End of the mission mission 4
Our approach Projection on Pattern (POP) UAV route Sweeping Pattern Local deviations => p D high => maximize p I and maintain p D high 5
The Flight Pattern Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Basic Sweeping Pattern (For static targets) repetitions of the pattern area to cover H flight backward pattern r D forward 7
Basic Sweeping Pattern (For static targets) backward Non-detected target T forward 8
Basic Sweeping Pattern (For static targets) backward T Detected target forward A target missed during the forward move is ensured be detected during the backward move 9
Enhanced Sweeping Pattern (For moving targets) backward Non-detected target T v T forward 10
Enhanced Sweeping Pattern (For moving targets) backward v T T forward Non-detected target 11
Enhanced Sweeping Pattern (For moving targets) backward Pattern narrowing T forward Detected target d T = f( α v ) with the velocity ratio between UAV and targets (see our paper for the expression of f and a sketch of proof) 12
Back to our example Situation at the beginning of the mission = 180 knots = 7 knots Critical targets 13
Back to our example Situation after executing one instance of the pattern Critical targets All the targets have been detected (and the targets T i at a distance to the pattern have been identified) 14
POP : Projection On Pattern Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Route Planning Problem T 2 T 1 How to investigate T 1 and T 2 while following the flight pattern ? 16
Targets projection Target T 1 Distance to the pattern 422.4 Projected P 13 target T 2 37.5 T 1 138.8 230.4 P 12 30.5 P 11 Arc length on the pattern 61.1 17
Targets projection Target T 1 422.4 T 2 37.5 T 1 138.8 230.4 30.5 P 1 61.1 18
Targets projection Target T 2 431.3 P 23 15.2 P 22 T 2 251.9 149.5 T 1 51.9 P 21 50.4 19
Targets projection Target T 2 431.3 P 2 15.2 T 2 251.9 149.5 T 1 51.9 50.4 20
Targets insertion 280.0 W 2 431.3 480.6 P 2 W 3 T 2 T 1 P 1 W 1 42.9 61.1 200.0 Flight pattern (copy) Projected targets P 1 W 1 P 2 W 2 W 3 Name ... 200.0 280.0 480.6 61.1 431.3 Arc length 21 increasing order of arc length
Targets insertion 280.0 W 2 431.3 480.6 P 2 W 3 T 2 T 1 P 1 W 1 42.9 61.1 200.0 Flight pattern (copy) Projected targets P 1 W 1 P 2 W 2 W 3 Name ... insertion preserving 200.0 280.0 480.6 61.1 431.3 Arc length order 22 increasing order of arc length
Targets insertion 280.0 W 2 431.3 480.6 P 2 W 3 T 2 T 1 P 1 W 1 42.9 61.1 200.0 Updated UAV route P 1 W 1 P 2 W 2 W 3 61.1 200.0 280.0 431.3 480.6 23
Targets insertion Postponed identification task 280.0 W 2 431.3 480.6 P 2 W 3 T 2 T 1 P 1 W 1 42.9 61.1 200.0 Updated UAV route P 1 W 1 P 2 W 2 W 3 61.1 200.0 280.0 431.3 480.6 24
Result on an entire mission p D = 100 % p I = 99.1 % POP Beginning of the End of the mission mission 25
Result on an entire mission The UAV route can be improved 3 enhancements proposed End of the mission 26
Enhancements Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Target move anticipation Without T is projected on the closest part of the pattern, regardless of its heading. T T resulting flight plan target projection 28
Target move anticipation Without Executed trajectory to identify T : B = 180 knots T = 6 knots A Elapsed time from A to B : 9753 s. 29
Target move anticipation With Estimated UAV position at interception Estimated target position time 2 at interception time 1 target projection resulting flight plan T 1 assuming a rectilinear motion of T 2 assuming a strict pattern following 30
Target move anticipation With Executed trajectory to identify T : B = 180 knots T = 6 knots A Elapsed time from A to B : 9638 s (previously 9753) Cost reduction : 1% 31
Amplified Pattern Narrowing Without Repeated deviations ⇒ UAV delay ⇒ Missed targets Missed target 32
Amplified Pattern Narrowing With No narrowing Normal narrowing Amplified narrowing Detected target (previously missed) 33
Amplified Pattern Narrowing With No narrowing Normal narrowing d T = f( α ' v ) Amplified narrowing Detected target (previously missed) α ' v = g( α v , D) ; g adjusts α v according to the UAV delay D ( see our paper for the expression of g ) 34
Local TSP Without Identifying targets by increasing arc length ⇒ oscillations W next Distance to W next : 296 NM 35
Local TSP Without Identifying targets by increasing arc length ⇒ oscillations W next TSP tour to improve Distance to W next : 296 NM 36
Local TSP With Distance to W next : 229 NM (previously 296) W next Tour improved by 2-opt Cost reduction : 23% 37
Demo Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Demo POP within the multi-agent platform APM(Robot) 39
Simulation results Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Experimental protocol Targets density L = 200 NM H = about 40 targets about 120 targets 200 NM Low density High density 41
Experimental protocol Parameter variants 42
Computation time Sensibility to density Low density High density 43
Computation time Sensibility to density Parameter variants Computation time under 10 ms in 99.9% of simulations Computation time mostly under 10 ms even with : - a high density of targets - all enhancements enabled Computation time under 10 ms in 98.1% of simulations 44
Highest values for p D + p I And corresponding parameter variants Low density High density 45
Values of v T and r I leading to p D + p I > 190% p D + p I > 190% p D + p I < 190% r I (NM) v T (knots) v T (knots) Low density High density 46
Conclusions and perspectives Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Presented work POP (Projection On Pattern) = Enhanced (i.e. narrowed) sweeping pattern + Target projections on this pattern + Target identification tasks ordered by increasing values of arc length 3 enhancements : ● target move anticipation ● amplified narrowing ● local TSP 48
Conclusions of simulation results The best combination of enhancements depends on the context. Enabling all enhancements does not necessarily lead to the best performances. High detection and identification performances can be obtained for reasonable targets speeds and identification range Computation time is mostly under 10ms, allowing on-the- fly replanning 49
Ongoing works • Extension to multiple cognitive UAVs • Dynamic identification tasks allocation ● Encouraging results with mTSP algorithms 50
End of talk Thanks for your attention. Slides available on my webpage. 51
Appendices Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Enhanced Sweeping Pattern Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Enhanced Sweeping Pattern Computation of d T Detected target 54
Enhanced Sweeping Pattern Computation of d T Detected target with 55
Enhanced Sweeping Pattern Enhanced Sweeping Pattern Computation of d T Root 2 Root 1 Root 1 Root 2 Keep the smallest root Downward move Upward move 56
Enhanced Sweeping Pattern Lower bound for Maximal narrowing No narrowing (basic pattern) The UAV must be faster than the targets 57
Amplified Pattern Narrowing Soulignac et al. UK PlanSIG'13 Edinburgh, 29 January 2014
Amplified Pattern Narrowing Principle overcost Arrival at T Estimated arrival at T' = T + v UAV l 3 l 2 overcost l 1 Situation 1 : Strict pattern following Situation 2 : Pattern following + local deviations The situation 2 is equivalent to the situation 1 with a UAV speed v' UAV < v UAV such that : l 1 + l 2 + l 3 v' UAV = T' 59
Recommend
More recommend
