This document presents a genetic algorithm approach for mission and safety planning with unmanned aerial vehicles (UAVs). It combines two existing genetic algorithm architectures - a hybrid genetic algorithm (HGA4m) for mission path planning and a multi-population genetic algorithm (MPGA4s) for replanning paths under critical situations. The proposed approach embeds both architectures into a single hardware system called MOSA (Mission Oriented Sensor Array) for mission supervision and IFA (In-Flight Awareness) for safety supervision. Computational results show the approach can find good solutions for both path planning and replanning within hardware limitations of onboard processors.

1. An Embedded System Architecture based on
Genetic Algorithms for Mission and Safety
Planning with UAV
Jesimar S. Arantes, Márcio S. Arantes, Claudio F. M. Toledo,
Onofre T. Júnior, Brian C. Williams
July – 2017
Jesimar S. Arantes, Márcio S. Arantes, Claudio F. M. Toledo, Onofre T. Júnior, Brian C. Williams (USP)GECCO 2017 July – 2017 1 / 27

2. Outline
1 Introduction
2 Problem Description
3 Methods
4 Computational Results
5 Conclusions
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 2 / 27

3. Introduction
This work addresses
Path planning for mission execution with UAV:
Hybrid Genetic Algorithm for mission (HGA4m) (Arantes et al. 2016).
Path replanning to land the UAV under critical situation:
Multi-Population Genetic Algorithm for security (MPGA4s) (Arantes et
al. 2015).
A combination of two architectures is proposed:
MOSA: Mission Oriented Sensor Array:
Supervision of mission systems;
Architecture create by (Figueira et al. 2013).
IFA: In-Flight Awareness:
Supervision of safety systems;
Architecture create by (Mattei et al. 2013).
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 3 / 27

4. Problem Description
End
airport
city
mountains
military
base
Start
(a)
Path planning problem with chance constraints and obstacle avoidance
Introduced by (Blackmore et al. 2011) and approached in (Arantes et
al. 2016)
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 4 / 27

5. Problem Description
End
airport
city
mountains
military
base
Start
bosk
bosk
bosk
bosk
plain
plain
(b)
Path replanning problem
Approached in (Arantes et al. 2015)
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 5 / 27

6. Problem Description
End
airport city
mountains
military
base
Start
bosk
bosk
bosk
plain
plain
bosk
(c)
Planning mission with system of safety
Embedded in the same hardware architecture
Approached in this work
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 6 / 27

7. Codification, Decodification and Solution
Codification ut:
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 7 / 27

8. Codification, Decodification and Solution
Codification ut:
Decodification FΨ:
xt+1 = FΨ(xt , ut )




px
t+1
py
t+1
vt+1
αt+1



 =




px
t + vt · cos(αt ) · ∆T + at · cos(αt ) · (∆T)2/2
py
t + vt · sen(αt ) · ∆T + at · sen(αt ) · (∆T)2/2
vt + at · ∆T − Fd
t
αt + εt · ∆T




J. S. Arantes et al. (USP) GECCO 2017 July – 2017 7 / 27

9. Codification, Decodification and Solution
Codification ut:
Decodification FΨ:
xt+1 = FΨ(xt , ut )




px
t+1
py
t+1
vt+1
αt+1



 =




px
t + vt · cos(αt ) · ∆T + at · cos(αt ) · (∆T)2/2
py
t + vt · sen(αt ) · ∆T + at · sen(αt ) · (∆T)2/2
vt + at · ∆T − Fd
t
αt + εt · ∆T




Solution xt:
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 7 / 27

10. Methods
Camera
Sensors
UAV
Actuators
GPSMotor
Servo Motor
Processor
AutoPilot
Follow
Waypoint
IFA System
Data
Acquisition
Sensor
Fusion
Security
Control
Decision
Making
Communication
Control
Path
Re-planning
MPGA4s [1]
MOSA System
Data
Acquisition
Sensor
Fusion
Mission
Control
Camera
Control
Communication
Control
Path
Planning
HGA4m [2]
embedded
embedded
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 8 / 27

11. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 9 / 27

12. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 10 / 27

13. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 11 / 27

14. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 12 / 27

15. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 13 / 27

16. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 14 / 27

17. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 15 / 27

18. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 16 / 27

19. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 17 / 27

20. Problem Description
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 18 / 27

21. Computational Results
Table 1: Settings used in the HGA4m and MPGA4s method.
Parameters Value HGA4m Value MPGA4s
number of populations 3 3
population size 3x13 3x13
crossover rate 5 0.5
mutation rate 0.7 0.75
stopping criterion 10 sec 1 sec
MigrationMigration
Migration
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 19 / 27

22. Computational Results
HGA4m
We evaluated 40 artificial maps in total
Stopping criterion 10 seconds
MPGA4s
We evaluated 60 artificial maps in total
We evaluated 4 critical situations
Stopping criterion 1 second
PC i5 Intel Edison
Frequency 1.8 GHz 500 MHz
Memory RAM 4 GB 1 GB
Operating System Linux - Ubuntu Linux - Yocto
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 20 / 27

23. Computational Results
HGA4m Method
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
100
1000
10000
100000
1000000
10000000
PC i5
Edison
Instances
TotalEvaluations
Figure 1: Number of evaluations by instance for path planning
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 21 / 27

24. Computational Results
HGA4m Method
8
10
12
14
16
18
20
PC i5
Edison
Instances
PathLenght
Figure 2: Path length by instance for path planning
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 22 / 27

25. Computational Results
MPGA4s Method
1
10
100
1000
10000
100000
Edison
PC i5
Instances
NumberofEvaluations
Figure 3: Number of evaluations by instance for path replanning.
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 23 / 27

26. Computational Results
MPGA4s Method
Edison
I1 1 1 2 1 1 1 2 1 1 1
PCi5
I1 1 1 2 1 1 1 2 1 1 1
I2 1 1 1 1 1 1 1 1 1 1
I2 1 1 1 1 1 1 1 1 1 1
I3 1 1 1 1 2 2 1 2 2 1
I3 2 2 2 2 1 1 2 1 1 2
I4 1 1 1 1 1 1 1 1 2 1
I4 1 1 1 1 1 1 1 1 1 1
I5 1 1 2 1 2 2 1 2 2 1
I5 1 1 2 1 2 1 1 2 2 1
I6 1 1 1 1 2 1 1 2 1 3
I6 1 1 1 1 2 1 1 2 1 3
I1 1 1 1 1 1 1 1 1 1 1
I1 1 1 1 1 1 1 1 1 1 1
I2 1 1 1 1 1 1 1 1 1 1
I2 1 1 1 1 1 1 1 1 1 1
I3 1 1 1 1 1 1 1 1 1 1
I3 1 1 1 1 1 1 1 1 1 1
I4 2 1 1 1 1 1 1 1 1 2
I4 1 1 1 1 1 1 1 1 1 1
I5 1 1 1 1 1 1 1 1 2 1
I5 1 1 1 1 1 1 1 1 1 1
I6 1 1 1 2 2 3 1 1 1 3
I6 1 1 1 1 1 1 1 1 1 3
I1 1 1 1 1 1 1 1 1 1 1
I1 1 1 1 1 1 1 1 1 1 1
I2 1 1 1 1 2 1 1 1 2 1
I2 1 1 1 1 3 1 1 1 1 1
I3 2 1 1 2 1 1 1 1 1 1
I3 1 1 1 1 1 1 1 1 1 1
I4 1 1 1 2 1 2 1 1 1 1
I4 1 1 1 1 1 1 1 1 1 1
I5 1 1 3 1 1 2 2 1 2 1
I5 1 1 3 1 1 1 1 1 2 1
I6 1 1 3 3 2 2 1 2 1 3
I6 1 1 3 3 1 1 1 1 1 3
I1 2 1 1 1 3 1 1 1 1 1
I1 2 1 1 1 3 1 1 1 1 1
I2 1 3 1 1 1 1 1 2 1 1
I2 1 3 1 1 1 1 1 1 1 1
I3 1 2 3 1 2 2 2 1 1 1
I3 1 1 3 1 2 2 1 1 1 1
I4 1 1 1 1 1 1 2 2 1 2
I4 1 1 1 1 1 1 1 1 1 3
I5 1 1 1 1 2 1 1 1 3 1
I5 1 1 1 1 2 1 1 1 3 1
I6 2 1 2 3 3 1 1 3 3 3
I6 1 1 1 1 1 1 1 3 3 3
Maps 1 2 3 4 5 6 7 8 9 10 Maps 1 2 3 4 5 6 7 8 9 10
Ψs1
Ψb
Ψm
Legend: Bonus Region Remainder Region No-Fly Zone
Ψm
Ψb
Ψs2
Ψs2
Ψs1
Figure 4: Landing sites in both architectures for path replanning.
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 24 / 27

27. Computational Results
Video Simulation - SITL to validate routes.
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 25 / 27

28. Conclusions
Despite differences in processing, the quality of the solution was
similar.
This indicates the robustness of GAs to find good solutions despite
hardware limitations.
In this way, GA can be embedded and aid the decision making during
fully autonomous flights.
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 26 / 27

29. J. d. S. Arantes, M. d. S. Arantes, C. F. M. Toledo, and B. C.
Williams.
A multi-population genetic algorithm for uav path re-planning under
critical situation.
In Tools with Artificial Intelligence (ICTAI), 2015 IEEE 27th
International Conference on, pages 486–493. IEEE, 2015.
M. d. S. Arantes, J. d. S. Arantes, C. F. M. Toledo, and B. C.
Williams.
A hybrid multi-population genetic algorithm for uav path planning.
In Proceedings of the 2016 on Genetic and Evolutionary Computation
Conference, pages 853–860. ACM, 2016.
L. Blackmore, M. Ono, and B. C. Williams.
Chance-constrained optimal path planning with obstacles.
IEEE Press, 2011.
N. M. Figueira.
Arranjos de sensores orientados à missão para a geração automática de
mapas temáticos em VANTs.
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 26 / 27

30. Tese de doutorado, Universidade de Sao Paulo (USP), mar 2016.
Sao Carlos, SP.
A. L. P. Mattei.
Consciencia situacional em voo de sistemas aereos nao tripulados.
Tese de doutorado, Universidade de Sao Paulo (USP), ago 2015.
Sao Carlos, SP.
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 27 / 27

31. Acknowledgements
Email to contact:
jesimar.arantes@usp.br
{marcio, claudio, otj}@icmc.usp.br
williams@mit.edu
Thank You!
J. S. Arantes et al. (USP) GECCO 2017 July – 2017 27 / 27