Motor-Level N-MPC for Cooperative Active Perception With Multiple Heterogeneous UAVs

This letter introduces a cooperative control framework based on Nonlinear Model Predictive Control (NMPC) for solving an Active Information Acquisition problem (AIA) using a system of multiple multirotor UAVs equipped with onboard sensors. The observation task of the NMPC is a minimum-uncertainty pose estimation of a moving feature which is observed by the multi-UAV system, using a cooperative Kalman filter. The controller considers a full nonlinear model of the multirotors - including the motor-level actuation units and their real constraints in terms of maximum torque - and embeds the Kalman filter estimation uncertainty in its prediction. The framework allows and exploits heterogeneity in the actuation and sensing systems by considering a generic model of UAV - including both quadrotors and tilted-propeller multirotors - and a generic model of range-and-bearing sensor with arbitrary rate and field of view. The capability of the proposed framework to reduce the cooperative estimation uncertainty of a static or a moving feature, thus leading the system to optimal sensing configurations, is demonstrated through Gazebo simulations and real experiments. The software is provided open-source.

Automated summary

This letter introduces a cooperative control framework based on Nonlinear Model Predictive Control (NMPC) for solving an Active Information Acquisition problem (AIA) using a system of multiple multirotor UAVs equipped with onboard sensors. The framework considers a full nonlinear model of the multirotors and embeds the Kalman filter estimation uncertainty in its prediction. The capability of the proposed framework to reduce estimation uncertainty and achieve optimal sensing configurations is demonstrated through simulations and experiments.