Fast model predictive control on a smartphone-based flight controller

Abstract

Several attempts have been made to execute control of a system using only a smartphone’s processor running computationally inexpensive algorithms such as PID, LQR or H-inf controllers. This paper presents design and implementation of model predictive controllers on a smartphone using the numerical optimization framework CasADi. To evaluate this framework’s performance (and compare its results with those from a Java library JOptimizer’s deployment) the implemented model predictive control algorithm was subjected to simulations of running a quadrotor control system on a smartphone. It attained a tracking error of 0.0693 m. These evaluation results open the possibility of implementing more computationally expensive algorithms on a smartphone’s processor including online or real-time usage.