Quadrotor deployment for emergency response in smart cities: A robust MPC approach

We consider the deployment of quadrotors for surveying damages and providing aid after disasters and in emergencies. A central assumption in this work is that as a result of the disaster, information regarding potential obstacles is either not available or not reliable. We therefore provide path planning and object avoidance algorithms that rely only on local sensing of the environment. We introduce a Robust Model Predictive Control (RMPC) method subject to input and output constraints which guarantees that all the obstacles in the local perception map are avoided even in the presence of polytopic uncertainties. Through a series of relaxations, the RMPC problem is formulated as a set of Linear Matrix Inequalities (LMIs), whose solution is guaranteed to be a feasible point for the original RMPC problem. A major improvement resulting from the proposed method is that the number of LMI constraints grows linearly with the number of uncertainties, which makes the method applicable to systems with relatively large number of uncertainties.