Distributed cooperative MPC for aerial robots: a ROS 2 implementation

The significant decrease in manufacturing costs of hardware components for quadrotors has greatly encouraged research into the design of flight control algorithm for quadrotors, which has seen great growth in recent years. One of the key aspects of the research is the communication between the quadrotors. Nowadays it is considered essential that the quadrotors can communicate with each other. This feature allows numerous advantages: it is possible to generate a network capable of collaborating to solve complex tasks that single quadrotors would not be able to perform, or complete them in a shorter time. The objective of this thesis is the design of a distributed algorithm to control the navigation of a set of quadrotors flying through the same navigation space. A surveillance task has been chosen as a case study, where quadrotors are in charge of arranging themselves in order to protect a target from intruders. Each quadrotor needs to complete both a specific task assigned to it (prevent a certain intruders from reaching the target) and a task in common with the other quadrotors (make sure that the center of the drones coincides with the target and the quadrotors do not collide). With this goal in mind, the project starts with the design of the quadrotor model, controller and trajectories from scratch. Then a Distributed Model Predictive Control algorithm is designed ad hoc to control the navigation of quadrotors. One of the challenges in the creation of this algorithm is the adaptation of the control algorithm to the simultaneous use of Model Predictive Control (MPC) and Online Distributed Gradient Tracking (O-DGT). Indeed, the speed required for the optimization calculations led us to reformulate the MPC in order to make the calculations faster and thus satisfy the limits imposed by the chosen time-step. The proposed model is tested with numerical examples, analyzing a series of cases that allowed us to test different combinations of the developed algorithms.