Detect & Avoid. UAV Integration in the Lower Airspace Traffic. Cyril Allignol, Nicolas Barnier, Nicolas Durand & Éric Blond. ICRAT June 22, 2016

Transcription

1 Detect & Avoid UAV Integration in the Lower Airspace Traffic Cyril Allignol, Nicolas Barnier, Nicolas Durand & Éric Blond ICRAT June 22, 2016

2 Context & Objective Demand for civilian UAV operations increases (fire detection, river bed surveillance, parcel delivery...) Most UAVs operate at low altitudes, interfering with traffic in TMAs Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 1 / 24

3 Context & Objective Demand for civilian UAV operations increases (fire detection, river bed surveillance, parcel delivery...) Most UAVs operate at low altitudes, interfering with traffic in TMAs How to ensure separation between UAVs and traffic? Managed by ANSPs Delegated to both aircraft and UAV (TCAS-like) Handled by the UAV only Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 1 / 24

4 Context & Objective Demand for civilian UAV operations increases (fire detection, river bed surveillance, parcel delivery...) Most UAVs operate at low altitudes, interfering with traffic in TMAs How to ensure separation between UAVs and traffic? Managed by ANSPs Delegated to both aircraft and UAV (TCAS-like) Handled by the UAV only Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 1 / 24

5 Context & Objective Demand for civilian UAV operations increases (fire detection, river bed surveillance, parcel delivery...) Most UAVs operate at low altitudes, interfering with traffic in TMAs How to ensure separation between UAVs and traffic? Managed by ANSPs Delegated to both aircraft and UAV (TCAS-like) Handled by the UAV only Our solution A geometrical detect & avoid algorithm Validated by intensive simulation Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 1 / 24

6 Contents 1 Model 2 Experimental Setup 3 Results 4 Conclusion & Further Work Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 2 / 24

7 Geometrical algorithm Based on [van den Berg et al., 2011] n-body collision avoidance algorithm for robotics... with a few substantial differences only the UAV handles the avoidance maneuver only heading changes are considered speed ratio between aircraft and UAV might be high: Aircraft from 200 kn up to 450 kn UAV from 80 kn to 160 kn Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 3 / 24

8 Geometrical algorithm Based on [van den Berg et al., 2011] n-body collision avoidance algorithm for robotics... with a few substantial differences only the UAV handles the avoidance maneuver only heading changes are considered speed ratio between aircraft and UAV might be high: Aircraft from 200 kn up to 450 kn UAV from 80 kn to 160 kn Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 3 / 24

9 Conflict detection A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 4 / 24

10 Conflict detection d A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 4 / 24

11 Conflict detection d A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 4 / 24

12 Conflict detection r d A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 4 / 24

13 Conflict detection r d d/τ A B τ is the anticipation time Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 4 / 24

14 Conflict detection r d d/τ A B τ is the anticipation time Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 4 / 24

15 Resolution principle r A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 5 / 24

16 Resolution principle s c r A B s C is the escape vector Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 5 / 24

17 Resolution principle s c r A A B s C is the escape vector Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 5 / 24

18 Resolution principle s c r s C is the escape vector A B B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 5 / 24

19 Resolution principle s c r A s C is the escape vector A B B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 5 / 24

20 Model A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 6 / 24

21 Model B A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 6 / 24

22 Model A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 6 / 24

23 Model A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 6 / 24

24 Model A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 6 / 24

25 Multiple constraints C B A Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 7 / 24

26 Multiple constraints C B A Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 7 / 24

27 Multiple constraints C B A Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 7 / 24

28 Multiple constraints C B A Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 7 / 24

29 Multiple constraints C B A Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 7 / 24

30 Resolution strategies Closest A Try to stay as close as possible to planned trajectory Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 8 / 24

31 Resolution strategies Closest C osest A Try to stay as close as possible to planned trajectory Saturates the constraint Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 8 / 24

32 Resolution strategies Safest A Try to keep the most room for maneuver Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 9 / 24

33 Resolution strategies Safest A Try to keep the most room for maneuver Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 9 / 24

34 Resolution strategies Safest A Try to keep the most room for maneuver t rget Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 9 / 24

35 Resolution strategies Safest A Try to keep the most room for maneuver Longer-term view S ƒ est t rget Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 9 / 24

36 Contents 1 Model 2 Experimental Setup 3 Results 4 Conclusion & Further Work Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 10 / 24

37 Experimental Setup Traffic Terminal Maneuvering Areas in Bordeaux FIR, France 475 recorded trajectories Simulations Scenarios are built so that if no maneuver is issued, there is a collision Target separation distance: d = 3 NM Many sets of parameters for more than simulations Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 11 / 24

38 UAV Experimental Setup Speed: 80 kn & 160 kn Turn rate: 3 /s to 7 /s Six different missions Anticipation time: τ = 5 min Resolution every 10 s Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 12 / 24

39 Contents 1 Model 2 Experimental Setup 3 Results 4 Conclusion & Further Work Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 13 / 24

40 Resolution Example Scenario Closest Safest Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 14 / 24

41 Comparison of strategies Closest Safest 300 Number of scenarios Closest distance of approach (NM) Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 15 / 24

42 Comparison of strategies Closest Safest 300 Number of scenarios Closest distance of approach (NM) Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 15 / 24

43 Influence of UAV maneuverability kn closest 80 kn safest 160 kn closest 160 kn safest Number of airprox UAV turn rate ( /s) Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 16 / 24

44 Influence of UAV maneuverability kn closest 80 kn safest 160 kn closest 160 kn safest Mean angle deviation ( ) UAV turn rate ( /s) Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 17 / 24

45 Contents 1 Model 2 Experimental Setup 3 Results 4 Conclusion & Further Work Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 18 / 24

46 Conclusion Detect & Avoid geometrical algorithm Intended to UAVs avoiding surrounding aircraft Heading change maneuvers at constant speed Validated through intensive fast time simulation against recorded traffic in TMAs Two different strategies Trying to stay as close as possible to mission trajectory Providing a better safety level Maneuverability (speed and then turning capacity) is the key for an efficient collision avoidance Still a few conflicts remain (mainly with the least maneuverable configurations) that need analysis Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 19 / 24

47 Further Work Strategies Hybridization Closest Safest Hybrid Number of scenarios Closest distance of approach (NM) Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 20 / 24

48 Further Work Strategies Hybridization Closest Safest Hybrid Number of scenarios Closest distance of approach (NM) Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 20 / 24

49 Further Work Strategies Hybridization Closest Safest Hybrid Number of scenarios Closest distance of approach (NM) Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 20 / 24

50 Further Work Strategies Hybridization Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 20 / 24

51 Further Work Back to mission Number of scenarios Closest distance of approach (NM) Closest Safest Hybrid Maneuvers might lead the UAV far from its mission Need to compute a route back to mission How to detect end of conflict? Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 21 / 24

52 Further Work Aircraft Trajectory Prediction A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 22 / 24

53 Further Work Aircraft Trajectory Prediction A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 22 / 24

54 Further Work Aircraft Trajectory Prediction A B Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 22 / 24

55 Further Work Aircraft Trajectory Prediction A B Resolution over several time steps Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 22 / 24

56 Further Work Enhanced strategy to escape traps A Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 23 / 24

57 Further Work Enhanced strategy to escape traps A Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 23 / 24

58 Further Work Enhanced strategy to escape traps A t rget Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 23 / 24

59 Further Work Enhanced strategy to escape traps A S ƒ est t rget Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 23 / 24

60 Further Work Enhanced strategy to escape traps A S ƒ est? t rget Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 23 / 24

61 Time for questions ENAC Lab Allignol, Barnier, Durand & Blond Detect & Avoid for UAV Integration ICRAT 2016, Philadelphia 24 / 24