Many state-of-the-art motion planning algorithms for UAVs rely on a simple kinematic model of aircraft dynamics, which is often referred to as Dubins vehicle model. On the other hand, Commercial Off-The Shelf (COTS) autopilots do not use such a model, and often only provide simpler capabilities such as waypoint navigation, or, altitude, speed, and turn rate hold. In this paper, in order to eliminate this gap, we describe an approach designed for implementation of controllers desgined for Dubins vehicles on a class of COTS autopilots. Given an open loop controller for a Dubins vehicle together with corresponding Dubins trajectory, we present an approach that uses output tracking to implement the open loop controller using the COTS autopilot closed-loop interface. The autopilots are modeled as black boxes accepting different sets of user inputs under different modes of operation. By choosing inputs to autopilots based on observed outputs, the autopilots track a pre-designed Dubins vehicle trajectory. Experimental results comparing the Dubins trajectory to the autopilot trajectory are presented. The approach provides an efficient way for systematic implementation of cooperative control algorithms on unmanned air vehicles without sacrificing their correctness or performance. The end result are the algorithms that work well in practice on UAVs using COTS autopilots and at the same time can be analyzed theoretically.
展开▼
