UAVs for Humanitarian Missions: Autonomy and Reliability

Transcription

1 2014 IEEE GLOBAL HUMANITARIAN TECHNOLOGY CONFERENCE (GHTC) October 10-13, 2014, Silicon Valley, San Jose, California USA, UAVs for Humanitarian Missions: Autonomy and Reliability Tullio Joseph Tanzi 1,2, Ludovic Apvrille 1, Yves Roudier 3, Jean-Luc Dugelay 3 Presented by Neslihan Kose 1 Institut Mines-Telecom, Telecom ParisTech, LTCI CNRS, Sophia Antipolis cedex, France, {firstname.name}@telecom-paristech.fr 2 URSI France, Commission F 3 EURECOM, Sophia Antipolis cedex, France, {firstname.name}@eurecom.fr Institut Mines-Télécom

2 Outline Humanitarian Data Natural Disasters Mini Drones What is it? Potential Objectives for Humanitarian Drones People detection People classification and counting Contact and inform, Technical Challenges Current Limitations New Architecture Data Inputs and Processing Conclusion 2 Institut Mines-Télécom

3 Natural Disasters Source: EM-DAT: The OFDA/CRED International Disaster Database Université Catholique de Louvain, Brussels (Belgium); 3 Institut Mines-Télécom

4 Mini Drones (UAV) Since the Hewitt-Sperry automatic airplane (1917), Drone Ancestor. 4 Institut Mines-Télécom

5 Potential Objectives Communication and coordination: Address communication black-out through an alternative network Secure Data Sharing Terrain reconnaissance: Exploration of areas of interest Using appropriate sensors (Optical, Microwave, etc.), Search And Rescue (SAR) operations: People detection People categorization and counting Contact and inform, 5 Institut Mines-Télécom

6 Humanitarian Drone Challenges Autonomy of the Drone: Do not require specific skills for rescue teams Do not create additional work for rescue teams Scattered and fast moving victims Evolving terrain conditions Reliability of the Drone: Prevent additional casualties Operating in hostile environments Adapted to a low level of equipment maintenance Preventing drone hijacks Strong ethical and deontological aspects: Respect for disaster victims and relief team Controlled use of acquired data, 6 Institut Mines-Télécom

7 Drone Safety and Security Dependability: Understanding of the environment (collision avoidance) Limitation of the impact in case of unexpected return to the ground Real time deadlines Secure Design: Securing communications (authentication / encryption) Protection of acquired data Validation of embedded architecture Energy management: Batteries weight vs. flight duration Optimization of power consumption Compliance with the constraints of the new generation of batteries (LiPo, etc.) 7 Institut Mines-Télécom

8 Our Contributions New Architecture for Drone Spatial completeness Comprehensive coverage of the area for establishing an emergency terrain mapping Autonomous Navigation and 3D Reconstruction Detection, classification and counting Affected people supervision Individual and group movements Estimation of people speeds 8 Institut Mines-Télécom

9 Towards New Architectures Flight Control C 1 C 2 C n I/O Avionics Sensors UAV On-board RF Module Remote Control I/O Payload Sensors Ground control station Internal Processing Facility Energy Safety and Security Telemetry Communications Mission Planning Data storage and processing Autonomous Emergency Control Autonomous Flight Control Autonomous Sensors Control Avionics Sensors Engines Control Payload Sensors 9 Institut Mines-Télécom

10 Spatial Completeness Find a optimal drone trajectory to cover the entire disaster area: Simple strategy for plane areas Require more complex strategies for a rugged terrain area, to optimize energy consumption 10 Institut Mines-Télécom

11 Autonomous Navigation and 3D Reconstruction 3D Vision with a mono-vision camera: Dense reconstruction (Fig A) requires to stop the flight Sparse reconstruction (Fig B) combined with a corkscrew flight trajectory Videos: Figure A Figure B 11 Institut Mines-Télécom

12 People Detection and Tracking Conventional image processing techniques: Face: Haar + Adaboost Body: HOG + SVM etc. Tracking People: Follow the movements of people (Kalman, particles filtering, etc.) Next Phase: Extract human traits (age, height, etc.) 12 Institut Mines-Télécom

13 Detection and Control (1) Video based Detection Right, left, up, down, away, close,... Interface A2 A1 A3 Drone Cmds B3 B2 B1 13 Institut Mines-Télécom

14 Detection and Control (2) 14 11/10/2014 Institut Mines-Télécom Modèle de présentation Télécom ParisTech

15 Conclusion Design a "civilian" Drone Towards on-board intelligence: Flight autonomy Mission autonomy Safety and security since the design stage Multi-source data fusion (Lidar, Radar, IR, etc.) Regulatory obligations (FAA in USA, DGAC, CNIL in France, etc.) Privacy (European Directive 95/46/CE: private data protection) Usage and acceptability versus "operational capability" 15 Institut Mines-Télécom

16 Perspectives Links between communicating objects and human: Definition and design of embedded antennas intended for the detection of EM emission such as the ones produced by mobile phones (GSM, UMTS, etc.) and wireless networks (Wi-Fi, Bluetooth, etc.) Integrate complex sensors: Radar, Lidar (DEM) Ground Penetrating Radar (GPR) Prototype validation in real conditions 16 Institut Mines-Télécom

17 Thanks for your attention Tullio Joseph Tanzi 1,2, Ludovic Apvrille 1, Yves Roudier 3, Jean-Luc Dugelay 3 Presented by Neslihan Kose 1 Institut Mines-Telecom, Telecom ParisTech, LTCI CNRS, Sophia Antipolis cedex, France, {firstname.name}@telecom-paristech.fr 2 URSI France, Commission F 3 EURECOM, Sophia Antipolis cedex, France, {firstname.name}@eurecom.fr URSI Commission F 17 page 17 Institut Mines-Télécom