Abstract

Since several decades ago, unmanned aerial vehicles (UAVs) have attracted a great deal of attention in academic, industrial and military communities. Recently, multiple cooperative UAVs have been applied in various applications such as forest fire detection and fighting, search and exploration, environmental monitoring and surveillance.
The main objectives of this dissertation are to design novel algorithms for single quadrotor UAV trajectory tracking control and multiple UAVs for cooperative/formation control. Then, applying these algorithms in forest monitoring and fire detection application, where a group of detection UAVs is required to surround and track the fire perimeter for monitoring and observation mission. Furthermore, a new algorithm for fault-tolerant cooperative control (FTCC) is proposed, in order to mitigate potential UAV fault effect for reliable and safe mission completion. Finally, a fire fighting algorithm is developed for achieving minimum distances for forest fire UAVs to arrive at their assigned fire spots destinations.
A combination of sliding mode control (SMC) and linear quadratic regulator (LQR) is used to design a single quadrotor UAV controller, which is then used to design a formation controller of multiple UAVs. Moreover, another formation controller is designed based on SMC to achieve robust formation control against modeling uncertainties and disturbances.
Cooperative UAVs are applied in forest monitoring and fire detection application through three stages: search, confirmation and observation. UAVs are assigned to search for potential forest fires in a certain area, once a fire is detected and a fire alarm will be generated by one or more of the UAVs. The UAVs team then reconfigures its formation by following an elliptic fire perimeter, calculated by the ground station (GS) using a fire spread model. Afterward, the fire alarm confirmation stage begins and all UAVs start evenly distributed for surrounding the fire spot according to the UAVs number in the team. When the fire alarm is confirmed, the observation stage starts and UAVs continue tracking the fire along the fire perimeter. SMC is used to design a formation reconfigurable controller to switch between a predefined formation shape during the search stage, to a dynamic surrounding formation. This controller guarantees even distribution of UAVs surrounding the fire spots and the robustness against disturbances. In addition, task assignment is used with multiple fire spots and multiple UAVs teams in order to reduce the mission execution time. Moreover, the proposed control algorithms are implemented to a team of UAVs paired with a team of unmanned ground vehicles (UGVs), by using these UGVs as a take-off and landing platform in forest monitoring and fire detection application.
Meanwhile, UAVs may need to leave formation for refueling/recharging during the mission of search, confirmation and observation, or if a fault occurred during the mission due to fire flames, heat or UAV's internal fault sources. Therefore, an FTCC algorithm is designed based on the graph theory to mitigate the fault effect on mission completion, and ensure complete surrounding and data gathering of the fire spots using different fire sensors such as infrared cameras, charge-coupled devices (CCD) cameras and thermal cameras etc.
Afterward, data gathered during observation stage are processed in the GS, then dangerous fire spots coordinates are sent to the fire fighting UAVs. The leader UAV, the GS or both can perform the task assignment process using an auction-based or Hungarian algorithms to assign each UAV to a fire spot for deploying fire suppressant. Furthermore, a hybrid approach of control parametrization and time discretization (CPTD) and particle swarm optimization (PSO) is proposed to achieve minimum flight distance for each UAV to arrive at its destination, minimizing fuel/battery consumption. Since PSO cannot solve the continuous control inputs, CPTD is used to provide an approximate piecewise linearization of the control inputs. Thus, PSO can be adopted to achieve the global optimum solution.
Finally, the proposed algorithms are being implemented on single and multiple quadrotor UAVs in simulations. While, the leader-follower approach is used in cooperative control in a decentralized manner to avoid the disadvantages of centralization. Thereafter, the proposed algorithms are verified on a set of Qball-X4 quadrotor UAVs and QGV unmanned ground vehicles (UGVs) platforms in real-time experiments through different scenarios.