HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation
Theses

Planification du mouvement 3D en temps réel de drones autonomes dans des environnements dynamiques inconnus

Abstract : Drones are used in increasingly complex environments, which leads them to fly autonomously without human supervision. In this context, a number of challenges must be addressed. The first is that of perception. The drone must have an important observation of its environment to be able to move efficiently. It must have an important observation of its environment to be able to move efficiently. It must then use this information to move safely. While ensuring that the commands generated are feasible by the robot. Due to their small size, the embedded computing power is quite low. This reduces the complexity of the usable algorithms. Moreover, GPS localization is not always possible. It is therefore necessary to have a localization based on perception, which increases the computational load. This manuscript deals with the problem of trajectory planning of UAVs in dynamic 3-dimensional environments. The results consist in the generation of a safe 3D trajectory, incorporating dynamic constraints to ensure its feasibility by the UAV through a controller. The main contributions are: - Use of the complete 3D information of the environment for the generation of trajectories. - Prediction of obstacle displacement and its consideration in the avoidance algorithm. - Obstacle avoidance based on geometry, dynamic capabilities of the UAV and variability of the environment. - Continuity of control: the beginning of a new trajectory respects the real state of the UAV and the new trajectory depends on the previous one. - Code embeddability, computation time constraints respected with respect to the flight. Using the local perception of the drone as well as its dynamics and that of the environment, Bézier curves are used as trajectory. In a completely autonomous way, an optimization under various constraints is carried out to find the best curve at each moment. This curve respects the dynamics of the drone and its real state at each time while being a safe path. The use of the real state of the UAV and old curves ensures the continuity between two successive curves. By building the trajectory, the speed and acceleration profiles are also impacted by the optimization. Finally, the addition of the curve ensures a smooth and feasible trajectory for the drone.In addition to a safe distance to obstacles, a speed cone is used to improve avoidance by constraining the trajectory optimization. Each obstacle has a velocity property. Three types of controllers are presented to track the generated curves. Simulation results show statistically the efficiency of the developed approach in different realistic environments. The impact of different parameters on the resulting trajectories is also studied. Finally, different real tests are performed to validate the computational load and the feasibility of the system on an embedded computer.
Document type :
Theses
Complete list of metadata

https://tel.archives-ouvertes.fr/tel-03656405
Contributor : Abes Star :  Contact
Submitted on : Monday, May 2, 2022 - 10:30:15 AM
Last modification on : Tuesday, May 3, 2022 - 3:46:04 AM

File

2022LIMO0030.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03656405, version 1

Collections

Citation

Julien Margraff. Planification du mouvement 3D en temps réel de drones autonomes dans des environnements dynamiques inconnus. Automatique / Robotique. Université de Limoges, 2022. Français. ⟨NNT : 2022LIMO0030⟩. ⟨tel-03656405⟩

Share

Metrics

Record views

0

Files downloads

0