심해저 연약지반용 무한궤도 차량의 경로추종 제어

Abstract

In recent years, there has been growing interest in marine mineral resources due to a gradually shortage of land minerals. Deep seabed manganese nodules, one of the major marine mineral resources, are found on the seabed at a depth of 4,500m5,500m. The deep seabed is strongly cohesive and adhesive with very low shearing strength. In order to drive on such soft deep-seabed, contact pressure should be reduced by distributing the weight of a vehicle. Therefore, the pilot mining robot for mining manganese nodules secured the sufficient contact area by adopting a four-row tracked vehicle. Unlike the steering method of general vehicles, the tracked vehicle uses the speed difference of the tracks, i.e. skid steering. Since this method generates slip and shearing displacement, a precise control of vehicle velocity is prerequisite. This thesis is to propose a path tracking approach considering slip for tracked vehicle driven on the soft deep-seabed. It also introduces the results of the real sea tests to verify the proposed algorithms. To achieve the purposes, the followings are performed. First, the mechanical system and the control system of the self-propelled pilot mining robot are examined. The pilot mining robot was developed in consideration of the underwater environmental elements and one-fifth scale of commercial capacity. The environmental elements which affect the vehicle driving on the deep-seabed are subgrade reactions and fluid resistance. This thesis describes the system configuration and the underwater sensors to control the robot. Second, mathematical model and system identification algorithm for tracked system with slip are proposed. The tracked system of pilot mining robot is an unknown system that cannot identify output against input. Therefore, the model parameters should be identified by the approximate ones of actual system. In this thesis, a simple numerical model is obtained from step response and graphical method. Based on the identified parameters, PI gain is designed through widely known gain tuning formulae. The PI gain guarantees small overshoot and fast settling time for control response. Third, an underwater localization algorithm for the precise decision of underwater position is proposed. Common methods for underwater localization are underwater positioning system and dead-reckoning method. The underwater positioning system has long sampling period and position error within certain range. On the other hand, dead-reckoning has a short sampling period, but the errors are accumulated as time passes. Therefore, this thesis proposes the new localization algorithm through fusing those two methods by using indirect Kalman filter. Fourth, two geometric path tracking methods are presented i.e. Follow the carrot algorithm and Pure pursuit algorithm. Path tracking is a method of controlling to reduce position error and heading error between the robot and the reference path. This thesis examines the principles of two methods, analyzes the simulation results, and selects the appropriate algorithm for the pilot mining robot driving on soft soil. Finally, the results of verification tests on proposed algorithms are presented. The tests are carried out both in shallow sea and in deep sea. System identification, controller design and underwater localization algorithm for the pilot mining robot were investigated in shallow sea. Underwater localization algorithm is tested in deep sea where underwater positioning system generates more noise. Also in shallow sea, control parameters are selected via the steering characteristic test, and path tracking control is performed.