Modeling and control for payload stabilization of slung-load transportation system

Abstract

This dissertation undertakes the modeling and parameter identification and the development of a control method for a slung-load transportation system. The slung-load system covered in this study consists of a rotorcraft and a payload along with a cable that is connected between the two. Both ends of the cable are attached to attachment points on the rotorcraft and the payload. Therefore, the rotorcraft attitude and the payload attitude have coupled dynamics due to these attachment points. The modeling method which relies on the Udwadia-Kalaba equation (UKE), advantageous when used with multiple constraints, is employed here to model the slung-load system. Depending on model fidelity differences, three types of slung-load systems are modeled: a simple 2D slung-load system, a general slung-load system, and a slung-load system with a flexible cable. In studies involving parameter identification and control system design processes, it is necessary to select and use a suitable slung-load model to ensure the correct degree of fidelity of the model required in each case. To identify the parameters of the slung-load system, a parameter identification method that utilizes the modal characteristics of the flexible cable of the slung-load system is proposed. The proposed method estimates the length of the cable and the mass of the payload by means of a frequency analysis. When employing the slung-load transportation system, a payload stabilization controller which stably transports the payload is necessary. First, a controller that combines an input-shaping method (ISM) and a model-following control (MFC) method is proposed here. The advantage of the proposed controller is that the intended payload oscillation induced by the input-shaping method is not considered to be a disturbance which affects the model-following control. Next, a payload stabilization controller using the deep deterministic policy gradient (DDPG) method based on reinforcement learning is proposed. For this purpose, a reward function suitable for a slung-load control system is proposed. The proposed DDPG-based payload stabilization controller is advantageous in that the payload can be stabilized by a model-free machine learning procedure. Finally, a simulation program is employed to verify that the proposed payload stabilization controllers operate as intended and to analyze their performance capabilities. The simulation results show that several performance indicators of the proposed controllers are improved compared to existing controllers.