Robust Multi-Agent Collision Avoidance for Drones

Al Ahdab, Mohamad

4. term

Control and Automation, Master

2019

2019-06-06

144 pages

This thesis aims at developing a robust controller with a local path planner to allow Crazyflies 2.0 to follow trajectories generated from a global planner while handling disturbances and uncertain parameters with collision avoidance. First, a quaternion based model is developed for the Crazyflie 2.0. After that, Sliding mode controllers (SMC) are developed for both attitude and position control to be connected in a cascaded fashion. Simulation tests confirmed the robustness of the suggested method. Afterwards, an IMU driven quaternion based Error-State Extended Kalman Filter with biases estimation is developed and tested in simulation and shown to have estimations accurate enough for the controller. Subsequently, an attempt to implement the position SMC on the Crazyflie 2.0 is carried out. The SMC managed to stabilize the $z$-axis but not the $x$ and $y$ axis. Further investigation is needed to solve this problem. Nevertheless, a simple robustness test is carried out on the $z$-axis and the SMC showed good robustness results. Finally, an MPC based local path planner is developed to work with the position SMC with a Kalman filter to estimate its dynamics and use it as a model for the MPC. The proposed algorithm is simulated for multiple drones flying together without colliding with each other and with a stationary obstacle. This shows a potential for this strategy to be implemented in real life.

MPC ; SMC ; Collision Avoidance ; Path Planning ; Robustness ; Drone ; Quadcopter ; Crazyflie ; Kalman filter

Download
View record in AAU Student Projects