Nonlinear Predictive UAV-Elastic Tool Interaction Control in Real-time

This paper addresses the control problem of an unmanned aerial vehicle (UAV) endowed with a tool to physically contact with the environment. Deploying an arm attached on UAV to utilize an interaction, particularly in close proximity to the environment, can be a more demanding task. To conduct a feasible operation, a tailored elastic arm is designed to employ a centralized predictive interaction control by leveraging the nonlinear quadrotor UAV model. For the system's physical bounds and sets for the UAV and its tool, an analysis is given. In order to maintain a feasible interaction and generate smooth trajectories, flatness-based outputs are defined. Consequently, a centralized optimization algorithm respecting system constraints and nonlinearities for the quadrotor-tool interaction is developed. The proposed approach is validated through experiments for the push recovery and sliding on the ceiling cases in real-time. Due to the dimension of the nonlinear model, the computation time is heavier than one of the possible decomposed models. However, the designed centralized algorithm for the constrained optimization-based control is solved in an average of 2.3 ms in this implementation.