Identifiez-vous pour accéder à la partie privée du site ou le modifier. Déconnexion.

Research topics:

The chair activity is organized around three main research topics.

1. Control and energetic optimization of mini-UAVs:

The chair activity focuses on small UAVs, with a mass limited to a few kilograms. The power of such systems is limited compared to the aerodynamic forces resulting from wind effects. Due to this high sensitivity to wind, control algorithms that can ensure the flight stability in a large range of angles of attack are needed. Part of our work concerns the design of feedback control laws endowed with large stability domains and robustness to unmodeled dynamics (wind effects, uncertainties on the UAV's mass and/or geometry, etc). Another topic of interest concerns the energetic autonomy of small UAVs. VTOL UAVs (for Vertical Take-Off and Landing) are often used for inspection applications because their ability to perform stationary flight allows to take pictures close to the inspected structure. A typical example is provided by the quadrotor. These systems, however, are not very efficient for cruising flight because they do not possess large wings. Therefore, they cannot benefit from associated lift properties. On the other hand, airplanes' wings  allow for a good flight energetic efficiency at high speed but airplanes usually cannot perform stationnary flight. We work on the design of "convertible" UAVs, i.e., UAVs that can perform stationnary flight and also cruising flight with good energetic performance. This requires understanding the limiting factors of flight performance and designing new flying structures.

2. Sensor data processing and sensor-based control for teleoperated or fully autonomous flight

Flight control of UAVs requires different sensors. IMUs (Inertial Measurement Units) with accelerometers, gyrometers, and possibly magnetometers are typically used to control the UAVs attitude (i.e. it's orientation in space). This is the most crucial component of flight control and attitude control is already very usefull for teleoperation. For inspection applications, sensors that can provide some information about the position of the UAV relative to the environment are important. Especially, it matters to have information about the distance to obstacles. Such sensors, like cameras or laser range finders are often termed "exteroceptive". Fully exploiting these sensors on small UAVs is challenging. First, data processing is time-consuming because these sensors provide a lot of information. This is particularly true for vision sensors. On the other hand, computing power is limited due to the limited payload of small UAVs. We work on the design of data processing algorithms for exteroceptive sensors that can be implemented on small UAVs. A direction that we've been investigating is the fusion of exteroceptive sensors with IMUs in order to speed up and robustify classical data processing algorithms. Another research direction concerns sensor-based control. In some situations, sensor information is not sufficient to estimate the complete pose but it is still sufficient to address the control objective. For example, the 3d-position of the UAV is not required for a wall following application. Only the lateral distance to the wall is needed. Sensor-based control consists in a "direct" use of the sensor information in the control loop. In some applications it presents several advantages w.r.t. more conventional (cartesian-based) feedback control. We investigate sensor-based control for UAVs with exteroceptive sensors like cameras or laser range-finders. A difficulty is related to the underactuation of UAVs. This is an important difference with "traditional" sensor-based control which has been essentially developed for fully actuated systems (like robotic arms). Sensor-based control can be used either in both teleoperation modes and fully autonomous flight.

3. Motion planning and high-level control strategies

This third research direction can be viewed as an extension of the previous one, with exteroceptive sensors now used to elaborate more global motion plans while the motion control in previous case is purely local (reflex-type). High-level control strategies rely on localization and mapping capabilities and require representation of the environment. Like for the previous research topic, sensor data processing for localization is essential but other techniques must be used and developed for on-line mission adaptation. This research direction, which is not yet under development, will  be investigated from data processing and fusion algorithms  developed in the previous research topic.