An Edge Architecture Oriented Model Predictive Control Scheme for an Autonomous UAV Mission

Achilleas Santi Seisa,Sumeet Gajanan Satpute,Björn Lindqvist,George Nikolakopoulos

DOI: https://doi.org/10.1109/ISIE51582.2022.9831701

2022-09-17

Abstract:In this article the implementation of a controller and specifically of a Model Predictive Controller (MPC) on an Edge Computing device, for controlling the trajectory of an Unmanned Aerial Vehicle (UAV) model, is examined. MPC requires more computation power in comparison to other controllers, such as PID or LQR, since it use cost functions, optimization methods and iteratively predicts the output of the system and the control commands for some determined steps in the future (prediction horizon). Thus, the computation power required depends on the prediction horizon, the complexity of the cost functions and the optimization. The more steps determined for the horizon the more efficient the controller can be, but also more computation power is required. Since sometimes robots are not capable of managing all the computing process locally, it is important to offload some of the computing process from the robot to the cloud. But then some disadvantages may occur, such as latency and safety issues. Cloud computing may offer "infinity" computation power but the whole system suffers in latency. A solution to this is the use of Edge Computing, which will reduce time delays since the Edge device is much closer to the source of data. Moreover, by using the Edge we can offload the demanding controller from the UAV and set a longer prediction horizon and try to get a more efficient controller.

Systems and Control