Towards autonomous flight of an unmanned aerial system in plantation forests

This paper presents an Unmanned Aerial System (UAS) for precision forestry applications. The UAS is capable of performing fully autonomous flights under the canopies of an unknown plantation forest by avoiding obstacles without the use of GPS navigation. The navigation framework of this system consists of a velocity estimator and an obstacle avoidance controller. In particular, the velocity estimator estimates raw velocity measurements from the 2D laser scanner using the Point-to-Line Iterative Corresponding Point (PLICP) scan matching algorithm and then fuses the filtered raw velocity measurements with readings from other on-board sensors using an Extended Kalman Filter (EKF). A high-level obstacle avoidance controller is realized by creating a local artificial potential field near the UAS, based on the goal point and nearby obstacles. The performance of the UAS is quantified using indoor and outdoor tests. During the outdoor test, in which the UAS flew autonomously for approximately 28 m within a stand of trees, the minimum distance between the obstacles and UAS was 1.38 m.