PATH PLANNING BASED ON RRT-KINODYNAMIC WITH NONLINEAR MODEL PREDICTIVE CONTROL FOR HYBRID AUTONOMOUS UNDERWATER GLIDER

• Main Author: Siregar, Simon
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/84510
• Institution: Institut Teknologi Bandung

Indonesia, as an archipelagic nation, holds tremendous potential in the maritime sector. However, the vast maritime territory also poses serious threats, both concerning natural resource issues and territorial disputes. To address these threats, an effective surveillance system is necessary, one of which is through the development of underwater vehicles capable of conducting long-term surveillance. One suitable option for such vehicles is the Autonomous Underwater Glider (AUG)- Hybrid. AUGs have been recognized for a decade as efficient vehicles for exploration and operations. However, to enhance their capabilities, the addition of "Hybrid" components such as thrusters becomes a fitting choice as it can improve the vehicle's maneuverability and speed. With the addition of hybrid systems, the need for adaptive and efficient control systems in navigating complex underwater environments becomes increasingly important. Therefore, the development of adaptive and efficient control systems, as well as accurate path and route planning, becomes the main focus of this research. This dissertation aims to develop an integrated approach in the development of a hybrid autonomous underwater glider (HAUG), which includes four main stages. The first stage involves the construction of mathematical models of the vehicle and the underwater environment as the basis for further development. The second stage focuses on the creation of simulations and the implementation of actual vehicles based on the developed models. In the third stage, efficient and adaptive control systems are introduced using Nonlinear Model Predictive Control (NMPC) and Trapezoidal Proportional Integral Derivative (PID) methods to address navigation challenges in dynamic environments. Finally, the fourth stage introduces a new approach to path and routing planning that utilizes RRT* algorithm with kinodynamic and NMPC integration, aimed at enhancing the navigation and control capabilities of HAUG vehicles in exploring underwater environments more efficiently and adaptively. Performance evaluation is conducted through simulations and field experiments, showing significant improvements in the navigation and control capabilities of HAUG vehicles. The results of this dissertation include several important contributions. First, it produces dynamic and kinematic models of the HAUG, as well as energy consumption and environmental models. These models are built using the Mission Oriented Operating Suite - Interval Programming (MOOS-IvP) and Robot Operating System (ROS) application frameworks. Simulations show that the performance of HAUG is in line with its design, achieving a maximum speed of 2 m/s in Autonomous Underwater Vehicle (AUV) mode and 0.503 m/s in glider mode. Additionally, in glider mode, simulations show pitch angle stability at angles between 21° and 30° with the lowest energy at 21° angle. Furthermore, the research shows that the NMPC-based control system has lower energy consumption compared to the Trapezoidal PID method. During testing, the energy consumption using the Trapezoidal PID method reached 113.64 J, while the NMPC control system required only 91.75 J with angles produced between 19° and 25°. Although tested with the same target and time, the results showed that NMPC is more efficient with 19.2% lower energy consumption. Additionally, simulation results using RRT*-Kinodynamic demonstrated that the developed algorithm could reduce the vehicle's travel time by up to 9.7% and energy usage by up to 12.03% in the tested cases. Performance evaluation is implemented through a combination of simulations and experiments. The results show that the RRT-Kinodynamic approach enhances the vehicle's ability to adapt to changing environmental conditions. The integration of NMPC and the RRT-Kinodynamic algorithm allows HAUG to reach almost all waypoints with an average Cross-Track Error (CTE) of 0.71 meters from GPS sensor measurements and 1.03 meters from simulations against the desired waypoint path. The standard deviation of CTE is 0.45 meters for the GPS path and 0.49 meters for the simulated path, confirming the adaptive and dynamic capabilities of this approach in dealing with complex environmental conditions.