Model Predictive Control for Distributed Electric Propulsion of eVTOL Vehicles: A Preliminary Design

View Video Presentation: https://doi.org/10.2514/6.2022-0878.vid This paper presents a preliminary design and feasibility study of a model predictive control (MPC) framework suitable for future flight control in distributed electric propulsion of eVTOL (electric vertical takeoff and landing) and/or UAM (urban aerial mobility) vehicles. An integral part of the MPC is a quadratic programming (QP) based optimizer solver, which can be accelerated by high-performance computing hardware and used to generate optimal control actions based on real-time model prediction. Machine learning algorithms can potentially be leveraged and incorporated in the controller to learn the model and parameters from real-time operation data, in addition to the known physical models. This MPC solver targets the problem of minimizing a predefined quadratic objective function, subject to equality and inequality constraints, and solve the optimization problem using the first-order optimality conditions. To find the solution online, the Jacobian matrix and gradient vector are updated iteratively, and then sent to a linear solver which is used to solve a set of linearized equation within each of the Newton’s iterations in the QP solver. Finally, after the linear solver finds a solution to the linear equation set, the QP solver’s solution vector will be updated. In the proposed MPC design, the controller aims to minimize the deviations from the desired trajectories, subject to the dynamic mission profiles, while managing the attitude of the vehicle. Preliminary simulation results from a pilot study show that the MPC control system operates well in tracking the desired trajectories and keeping the vehicle’s attitude angles within limits, under dynamic conditions. Further work will be ongoing to implement the control scheme on hardware and integrate the learning capability into the MPC controller.