DESIGN AND IMPLEMENTATION OF PITCH AND DEPTH RATE CONTROL SYSTEM USING MODEL PREDICTIVE CONTROL WITH ARX MODEL ON AUTONOMOUS UNDERWATER GLIDER

• Main Author: Jeremy, Jason
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/80183
• Institution: Institut Teknologi Bandung

The underwater glider is a submersible vehicle used for ocean exploration. It moves by gliding, which involves ascending, and descending at specific pitch angles. Gliding movement is controlled using a moving mass actuator to manage the pitch angle and a buoyancy engine for vertical movement within the water. However, controlling the gliding motion faces challenges due to factors such as non-linear systems both internally and externally, system delays, coupling effects, system constraints, and the need for energy efficiency. This research aims to design a control system for managing an underwater glider, focusing on energy efficiency issues. The adopted control method is Model Predictive Control (MPC) with an AutoRegressive eXogenous input (ARX) model. MPC is chosen for its ability to generate optimal input signals while considering system constraints and energy efficiency simultaneously. The selection of MPC is supported by choosing a computationally lightweight model, and ARX is chosen as it captures system dynamics through relatively low-computation input and output. MPC with ARX is designed to control the pitch angle and depth rate of the glider. The depth rate is regulated using a trapezoidal motion profile, indirectly controlling the glider's depth. The ARX pitch model, configured with a first-order output and second-order input, yields an RMSE (Root Mean Square Error) of 2.6 degrees. Meanwhile, the ARX depth rate model, configured with a first-order output and second-order input, produces an RMSE of 8 mm/s. The prediction horizon for MPC is set at 15 seconds with a computational time of 0.48 seconds, utilizing simple adaptive weights that switch between 10^(-3) and 1 to adjust tracking error conditions. The designed MPC model demonstrates a sevenfold improvement in efficiency for the moving mass actuator and a twofold improvement for the buoyancy engine compared to a fixed weight. MPC ARX control implementation shows tracking capabilities, even with depth shifts and pitch oscillations around the setpoint. The average absolute errors in the tracking process are 0.27m, 0.017m/s, and 2.8 degrees for depth, depth rate, and pitch, respectively. iv Comparing to PID control, PID exhibits superior rise times in depth and pitch control compared to MPC. However, MPC excels in pitch control with smaller average absolute errors. In terms of energy efficiency, MPC produces more optimal inputs, resulting in a 27.8% energy savings compared to PID.