Towards Autonomous Flights of an Unmanned Aerial System (UAV) in Plantation Forests

Abstract

Safe and autonomous operation of an Unmanned Aerial Vehicle (UAV) below the canopy of a plantation forest is a challenging yet worthwhile task for precise forest management. The research described in this thesis demonstrates the development and performance evaluation of a UAV for autonomous below-canopy flights. The focus of this research was on obstacle avoidance and motion estimation of the UAV, which are two critical tasks in a successful autonomous operation. A navigation block diagram for autonomous below-canopy flights was proposed in this research and the environment in a typical New Zealand plantation forest was studied. A below-canopy flight simulation environment, which includes two forest models, was developed based on existing software. Two obstacle avoidance (OA) controllers were implemented for autonomous below-canopy operation: a Vector Field Histogram (VFH) OA controller and an Artificial Potential Field (APF) OA controller. The implementations of the two OA controllers are slightly different from conventional implementations for simplicity and better reliability in cluttered natural environments. The performance of the two OA controllers was verified in the simulation environment. A velocity estimator was designed for the UAV. Raw velocities are estimated using the Point-to-Line Iterative Corresponding Point (PLICP) scan matching algorithm. An Extended Kalman Filter (EKF) is used to fuse the estimated raw velocities and readings from other sensors to provide velocity measurements with better reliability and a higher update rate. The performance of the developed velocity estimator and the UAV system was quantified in a series of indoor and outdoor experiments. The experimental results show that the maximum RMS error in the velocity estimation was 0.0768 m/s and the developed UAV successfully achieved autonomous flights both in the indoor test environment and in a plantation forest. Factors which affect the performance of the velocity estimator or the OA controllers are