PhD: Grenoble Institute of Technology

Contributed by: Hassen Fourati, hassen.fourati@gipsa-lab.grenoble-inp.fr

PhD Position: Multi-sensors estimation and fault-tolerant real-time control of unmanned aerial vehicle

Supervisor: Daniel Simon (Junior Researcher INRIA, HDR), daniel.simon@inria.fr, +33 (0)4 76 61 53 28

Co-supervisor: Dr. Hassen Fourati (Associate Professor, UJF), hassen.fourati@gipsa-lab.fr, +33 (0)4 76 82 64 25

Laboratory: GIPSA-lab (Grenoble Images Parole Signal Automatique)

www.gipsa-lab.fr, UMR 5216 CNRS, Joseph Fourier University-Grenoble I, Grenoble Institute of Technology (INPG)

Research team: NeCS (GIPSA/INRIA – National Institute for Research in Computer Science and Control) necs.inrialpes.fr.

Dates: October 2012 – September 2015

Candidate profile: Preferably with a Master's degree in Automatic or Robotics.

Type of funding: Doctoral contract. Gross income: 1 676,55 euros per month.

Application: CV, cover letter for application, certificates and notes, recommendation letter.

Project description:

The Unmanned Aerial Vehicles (UAVs) are aircrafts capable of flying and realizing a mission without people on board. Originally developed as part of military activities, there is now great potential for civilian activities (monitoring, mapping...). Many constraints remain unresolved for the use of UAV in public space. Among the key aspects to be tackled are embedded decision autonomy, the ability to perceive the environment at all times, safety and dependability. These characteristics will in future be subject to a certification process under development, but current UAVs still suffering from a lack of robustness and autonomy.

The flight control systems (combining Inertial Measurement Unit, sensors, motors, actuators...) provide stability and control functions and navigation of UAVs. For example, an inertial unit, is necessary to calculate the attitude of the UAV in flight, is often composed mainly of triads of MEMS (Micro-Electro-Mechanical Systems) accelerometers, magnetometers and gyroscopes. These sensors are prone to defaults (magnetic disturbances, biases...) which subsequently affects the control system (the calculation of the attitude of the UAV and its stability in flight).

Many approaches for the attitude estimation are still unreliable and often drift over time. Preliminary works have been developed in [1], [2] and possible improvements in the strategies of robust estimation, combining inertial, magnetic data and / or GPS, are still possible. On the other hand, the purely tele-operated control of an UAV, especially in a cluttered environment is a delicate task that must be facilitated by the autonomous execution of local actions such as for monitoring or avoidance of obstacles.

In this context, the proposed thesis can be developed in two parts with the following objectives:

1. In the first part of the thesis, we will focus on the problem of attitude estimation (3D spatial orientation) of the UAV. This information is often necessary for navigation (stabilization of the UAV) in flight condition. The results of previous works in literature at this level are encouraging but significant discrepancies are still observed in the attitude estimation in case of sudden and accelerated movements of the UAV

[3]. To solve this problem, we propose new data fusion approaches based on complementary filtering for the states estimation and observation by combining inertial measurements (accelerometers and gyroscopes) and magnetic measurements (magnetometers) and without resorting at each time to GPS and velocity measurements.

The proposed methods until now for the attitude estimation in the case of UAVs are based on the triad of sensors mentioned latter; we will search if it is possible at a final step to overcome the gyro data and its intrinsic bias. In this case the method will be reduced to the use of accelerometer and magnetometer.

2. In the second part of the thesis, we will develop some fault-tolerant controls (sensors and actuators defaults, but also real-time execution defaults and / or loss of connection with the master station) using the sensors measurements and the available execution resources. The approaches we propose will be based on the design of observers for the isolation and estimation of defaults, as well as the design and implementation of robust control laws and flexible real-time scheduling as outlined in [4].

Bibliography:

[1] H. Fourati, N. Manamanni, L. Afilal, and Y. Handrich. A nonlinear filtering approach for the attitude and Body Acceleration estimation based on inertial and magnetic sensors: Bio-logging application. IEEE Sensors Journal, vol. 11, no. 1, pp. 233-244, January 2011.

[2] H. Fourati, N. Manamanni, L. Afilal, and Y. Handrich. Posture and body acceleration tracking by inertial and magnetic sensing: Application in behavioural analysis of free-ranging animals. Biomedical Signal Processing & Control (BSPC), Elsevier, vol. 6, no. 1, pp. 94-104, January 2011.

[3] R. Mahony, T. Hamel, and J. M. Pflimlin. Nonlinear Complementary Filters on the Special     Orthogonal Group. IEEE Transactions on Automatic Control, vol. 53, no. 5, pp. 1203-1218.

[4] C. Berbra, S. Gentil, S. Lesecq and D. Simon. Control and Diagnosis for an Unmanned Aerial Vehicle, in: Co-design approaches for dependable networked control systems, ISTE Wiley, January 2010.