부족구동 역학 시스템의 비선형 적응제어

박문수 아주대학교 2007 국내박사

This dissertation is devoted to a nonlinear adaptive stabilization and tracking control of a class of underactuated mechanical systems. Underactuated systems are mechanical control systems with fewer controls than the number of configuration variables. Control of underactuated systems has been one of the active research fields due to their broad range of applications in robotics, aerospace vehicles, and marine vehicle. For example, underactuated systems in real-life include flexible-link robots, mobile robots, walking robots, robots on mobile platforms, cars, locomotive systems, snake-type and swimming robots, acrobatic robots, aircraft, spacecraft, helicopters, satellites, surface vessels, and underwater vehicles. Recent survey shows that the control of general underactuated mechanical systems is still one of the major open problems. The objective of this dissertation is to develop several methods for the configuration stabilization and trajectory tracking of a class of underacuated mechanical systems. The configuration stabilization is addressed for non-minimum phase underactuated systems (e.g., inverted pendulum) having unstable zero dynamics. In this dissertation, the scope of configuration stabilization is focused on the global or semi-global stabilization as well as orbital stabilization (i.e., stabilization of limit cycle). On the other hand, the trajectory tracking is considered for both minimum and non-minimum phase underactuated systems. For minimum phase underactuated systems, decoupling control based on collocated partial feedback linearization is adopted, which includes additive variable structure control law (VSC) for the asymptotical stabilization of unactuated subsystems. On the other hand, a novel partial feedback linearization (so called ”Lagrangian preserving partial feedback linearization”) is developed and utilized for the decoupling control of non-minimum phase underacuated systems. In addition, an adaptive schem based on fuzzy systems is proposed, to cope with systems uncertainties including system parameter uncertainties, external disturbances, and actuator nonlinarities for more practical issues. The main results of this dissertation are summaried as follows. First, semi-global and orbital stabilizations are presented by using backstepping technique, sliding mode control (SMC), and a novel partial feedback linearization. In backstepping based approach, we define an error state that utilizes momentum conjugate to unactuated variables which make it easy to apply standard backstepping approach. In stability analysis for the whole systesm, a condition that ensures semi-global and almost global domain of attraction is provided. In SMC based approach, we firstly decompose an underacuated mechanical system into actuated and unactuated subsystems, followed by a design of individual control law for each subsystem. Semi-global stabilizations of equilibrium and relative equilibrium point, as well as orbital stabilization around equilibrium are demonstrated through an illustrative numerical example. In a novel partial feedback linearization based approach, we define a weakly (or marginally) minimum phase output of a non-minimum phase underactuated mechanical system. For this new output, partially feedback linearizing control is designed, which renders the integral of closed-loop zero dynamics be an energy-like function having Lagrangian form. This energy-like function has its maximum value at equilibrium point and differs from a Lyapunov function in the sense that it is not positive or negative definite. Then, two methods for the asymptotic stabilization of equilibrium are propsed; the first method utilizes a nonlinear dissipation control in such a way that the energy-like function converges to its maximum value (energy-maximizing control); second method considers a modified version of the energy-like function as a Lyapunov fucntion. Second, trajectory tracking of underactuated mechanical systems is based on the decoupling control and partial feedback linearization. For minimum pahse systems (e.g., overhead cranes), we adopt collocated partial feedback linearization and employ a nonlinear dissipation control law to guarantee the asymptotic stability of unactuated subsystem. To demonstrate the proposed method, we perform numerical simulations and experiments for an overhead crane without hoisting motion. For non-minimum phase systems (e.g., inverted pendulum), we utilize the Lagrangian preserving partial feedback linearization and design trajectory tracking control law for the redefined output. By introducing a nonlinear dissipation control, asymptotic and bounded trajectory trakcing is accomplished. Finally, fuzzy based adaptive scheme to deal with systems uncertainties is proposed, where fuzzy system is utilized to estimate two lumped uncertainties for each actuated and unactuated subsystem. In this scheme, we treat uncertainties as a lumped one which is coupled with each other. Thus, we can estimate both uncertainties for actuated subsystem and unactuated subsystem with one uncertainty observer for the unactuated subsystem.

Lightweight underactuated bionic hand driven by shape memory alloy actuator with locking system

Hwang, Seongtaek Sungkyunkwan university 2021 국내석사

This study was conducted with the aim of producing ultra-light bionic hand driven by Shape Memory Alloy actuators with locking system. The lack of space due to the use of conventional SMA alloy wires has placed many restrictions on implementing various degrees of freedom for robot hand and arms. And due to the weak force of the actuator, the robot hand did not carry high load and did not fulfill its hand function. To solve this problem, a SMA spring actuator with a large strain was introduced and a locking system was applied to design a finger that could withstand high loads and, using this, a ultra-light bionic handle to lift high loads was produced. Chapter 1 introduced active research through artificial muscles due to the development of artificial muscles to replace motors in the future, and introduced SMA that are attracting attention for their outstanding performance. Robot hands and arms driven by SMA also are introduced and considered the volume, weight, and load of the system that they currently have. Based on these problems, the problem was defined and solutions were presented. Chapter 2 deals with and verifies the design of the locking system for holding high loads and the finger design to apply them. The locking system consisted of ratchet and switch. Therefore, when the shape memory alloys drive is pulled, pawl is lifted and moves unrestricted, but when it moves in the opposite direction, pawl does not rise, so it is stuck with the ratchet and maintains its position. However, for finger application, bidirectional operation was required to enable upward and downward movement of pawl with the movement of retracted mechanism and SMA spring actuator. Then, the underactuated mecchanism with tendon was applied to fingers and combined it with the locking system to develop a finger that can lift high loads. Finally, the finger was confirmed that it can lift high loads through weight test. In Chapter 3, the ultra-light bionic hand was designed based on the fingers applied with the locking system that withstand the high loads. Designed with a size similar to a human hand, the Bionic Hand is designed with a 3D printer and weighs very light, with a total mass of 223.3g. Also, it has a total of 10 degrees of freedom, with each finger capable of two-way operation, and a locking system has been applied to the index finger and ring finger. This allows each finger to move freely, grab a variety of objects, and lift 4.2 kg, about 18 times the weight of the hand, through a two-finger locking system. 본 연구는 형상기억합금 구동기를 이용한 락킹 시스템이 적용된 초경량 바이오닉 로봇 핸드 제작을 목표로 진행되었다. 기존의 형상기억합금 와이어의 사용으로 인한 공간의 부족은 로봇 암의 다양한 자유도를 구현하는데 많은 제약을 가져다주었다. 그리고 구동기의 약한 힘으로 로봇 핸드는 높은 하중을 들지 못하여 손의 역할을 다하지 못하였다. 이러한 문제를 해결하고자 큰 수축률을 가지는 형상기억합금 스프링을 도입하고 락킹 시스템을 적용하여 고하중을 견딜 수 있는 손가락을 설계하고 이를 이용하여 고하중을 들 수 있는 초경량 바이오닉 핸드를 제작하였다. 제1장은 앞으로 모터를 대신할 인공근육의 발전으로 인공근육을 통한 활발한 연구가 진행됨을 소개하며 그 중 뛰어난 성능으로 주목받고 있는 형상기억합금을 집중적으로 소개하였다. 그리고 형상기억합금으로 구동되는 로봇 핸드와 암을 소개하고 현재 이들이 가지는 문제점인 시스템의 부피와 무게, 그리고 들 수 있는 하중에 대하여 고찰하였다. 이러한 문제를 바탕으로 문제를 정의하고 해결방안을 제시하였다. 제2장에서는 바이오닉 핸드가 고하중을 들기 위한 락킹 시스템의 설계와 이를 적용할 손가락의 설계를 다루고 이를 검증하였다. 락킹 시스템은 라쳇과 스위치로 구성되었다. 그래서 형상기억합금 구동기가 당길 때는 풀이 들리면서 제약을 받지 않고 움직이지만, 반대 방향으로 움직일 때는 풀이 올라가지 않아 라쳇과 걸리면서 위치를 유지하게 된다. 그러나 손가락에 적용하기 위해서는 양방향 구동이 돼야 하여 풀을 리트랙티드 구조와 형상기억합금 구동기의 움직임으로 상하 움직임이 가능하게 하였다. 그리고 텐던으로 움직이는 언더액츄에이티드 구조를 손가락에 적용하고 이를 락킹 시스템과 결합하여 고하중을 들 수 있는 손가락을 개발하였다. 최종적으로 고하중 실험을 통하여 손가락이 고하중을 견딜 수 있다는 것을 검증하였다. 제3장에서는 앞서 만든 고하중을 견디는 락킹 시스템이 적용된 손가락을 바탕으로 초경량 바이오닉 핸드를 설계하였다. 사람의 손과 유사한 크기로 설계된 바이오닉 핸드는 3D printer로 설계되어 무게가 매우 가벼워 총 223.3g의 질량을 가진다. 각각의 손가락이 양방향 구동이 가능하면서 총 10 자유도를 가지고 검지와 약지에 락킹 시스템이 적용되었다. 이를 통해 각각의 손가락을 자유롭게 움직일 수 있고 다양한 물체를 잡을 수 있으며 두 손가락의 락킹 시스템을 통해 손 무게의 약 18배에 달하는 4.2kg을 들어 올릴 수 있다.

Balancing and Transferring Controls of the Ridable Ballbot

PHAM DINH BA Kyung Hee University - Graduate School 2018 국내박사

This thesis focuses on the design and implementation of a ridable ballbot and application of control methodology to control the ridable ballbot. The control methods, which include linear double-loop control and the nonlinear control techniques, such as partial feedback linearization (PFL) and sliding mode control (SMC), are designed based on two versions of the mathematical models of the ridable ballbot system: one describes the two-dimensional (2D) motions of ridable ballbot and the other is extended for its two and half-dimensional (2.5D) motions. The performance and effectiveness of the proposed control methods and accuracy of the mathematical models are demonstrated by both simulation and experiment. The simulation is realized in the PC using Simulink Matlab. The real ridable ballbot is installed for experiment with Visual Studio C++. Generally, a ridable ballbot has a body that balances on top of a ball and forms an underactuated system. The robot has more degrees of freedom (DOF) than independent control inputs. The ridable ballbot accelerates by leaning, but cannot directly control its tilt angles. The body is kept upright at approximately its unstable equilibrium point by only maneuvering the ball, similar to controlling an inverted pendulum. The controllers are first designed for the planar ridable ballbot then improved for the spatial motion of ridable ballbot system. The proposed control scheme concurrently implements three main tasks: balancing, transferring, and rotating a ridable ballbot on the floor. Based on the application of multi-loop control, linear quadratic regulator (LQR), PFL, and SMC techniques, this study proposes one linear optimal double-loop controller and six types of nonlinear controllers represented as follows. (i) Linear double-loop control system with LQR outer-loop and proportional-integral(PI)-plus-feedforward inner-loop controllers. This study proposes a linear double-loop control system based on planar ridable ballbot models. These planar models are linearized around an equilibrium point where the ridable ballbot is in an upright position and stationary. The double-loop control system includes two controllers: one is a PI-plus-feedforward inner-loop controller and the other is an LQR outer-loop controller. The double-loop control system allows the ridable ballbot to overcome the hindrance caused by its heavy weight and bulk and eliminate most of the uncertainties of system parameters, modeling, friction and external disturbances. (ii) The PFL based controller for planar ridable ballbot motion and its extended version for spatial motion. The PFL controller is designed to control balancing and transferring of the ridable ballbot. The planar dynamics of the ridable ballbot is formulated with highly nonlinear second order differential equations. This planar dynamics involves the ball state and body state driven by only one torque input. First, its system dynamics should be divided into ball subsystem model and body subsystem model. Next, two PFL laws are proposed; one is determined by the ball subsystem and the remaining law is got from the feedback linearization of the body subsystem. Considering the body state as the primary system output, the PFL control input is proposed to balance and track the position of the ball to the desired reference trajectory. Because this nonlinear control law does not guarantee the convergence of the overall state variables, the structure of the controller should be modified to satisfy the stability condition of the full state variables. The modification is a linear combination of two components: one is received from feedback linearization of the ball dynamics and body dynamics. Then the proposed controller asymptotically pulls all system responses to the desired values. Improving the aforementioned proposed controller for the spatial ridable ballbot motion, another nonlinear controller is considered for the robot in the case of the most complicated tasks in which the ball, three omnidirectional wheels, and body motions are simultaneously operated. The control structure is designed based on the nonlinear feedback of the ball and body states. First, system dynamics of ridable ballbot composed of four highly nonlinear second-order differential equations are created. Because the ridable ballbot is a second-order underactuated system with non-holonomic velocity constraints and input coupling case, its system dynamics should be separated into ball and body subsystems. Next, the two sub-systems are linearized by using the nonlinear feedback technique. Afterward, considering states as system outputs, a nonlinear controller of the ball subsystem is proposed to track output positions of the ball on the floor to the desired path while another nonlinear controller of the body subsystem is proposed to keep the body at the vertical position. The stability analysis of the two subsystems and the overall system is given to prove the convergence of the system states. Finally, the performance of the proposed nonlinear control method is simulated and experimentally tested for self-balancing and station-keeping, transferring, and robustness against external disturbances. (iii) The hierarchical SMC based controller for the planar motion of the ridable ballbot and its extended version for the spatial motion. SMC is a robust technique applied widely in nonlinear systems to implement and insensitive with external disturbances and uncertainties of system parameters, friction, and modeling. It is especially suitable for underactuated systems, such as ballbot system. The goal of SMC approach is to determine an asymptotically stable surface so that all state trajectories are attracted to this surface and slide along until reaching their desired values. As mentioned, the ridable ballbot is an underactuated system containing uncertain components, such as system parameters, friction factors, and model. Therefore, SMC technique is a proper selection in this study. Two hierarchical SMC schemes are discussed. One is for the planar motion of the ridable ballbot system and the other is improved for the spatial ridable ballbot system. Based on this physical structure, the hierarchical structure of the sliding mode surfaces (SMS) is designed as follows. At first, the SMS of every subsystem is defined. Then the SMS of one subsystem is defined as the first layer SMS. The first layer SMS is used to construct the second layer SMS with the SMSs of another subsystem. This process continues till the SMSs of the entire subsystems are included. According to the hierarchical structure, the total control law is deduced by the Lyapunov theorem. In theory, the asymptotic stability of the entire system of SMSs is proven and the parameter boundaries of the subsystem SMSs are given. (iv) PFL and SMC combination based control scheme for the planar and spatial motion of the ridable ballbot. An improved nonlinear control scheme for the planar motion of the ridable ballbot is proposed applying a combination of two nonlinear control techniques. The first step in combined control design is to decompose the ridable ballbot system into ball and body subsystems. Next, using the PFL technique, the body subsystem control is constructed to control balancing of the robot. The ball transferring control is designed based on the SMC technique and ball subsystem. In the final step, the balancing and transferring control laws are merged to stabilize overall state variables of the ridable ballbot system. Improving the aforementioned controller for the spatial motion of the ridable ballbot, the control scheme is divided into two parts in which each of them implements a particular duty. PFL controls the balance of the body while SMC controls the position of the ball on the floor. A combined nonlinear controller is proposed by integrating PFL based scheme into SMC based structure.

Wire rope contact model and crane control method for the advanced simulation and automation of block erection

이혜원 서울대학교 대학원 2020 국내박사

선박 및 해양구조물은 여러 개의 블록을 탑재하여 건조된다. 블록의 탑재 작업을 수행 시 안전을 위해 사전에 물리 기반 시뮬레이션을 통해 발생할 수 있는 위험성을 미리 파악하는 것이 요구된다. 또한, 실제 작업의 제어는 수작업으로 이루어지는데, 보다 안전하고 효율적인 탑재 작업을 위해 크레인의 제어 자동화 방법에 대한 연구가 필요하다. 본 연구에서는 기존의 시뮬레이션 에서 모사할 수 없었던 와이어 로프와 블록 간의 간섭이나 해상 크레인과 계류 시스템 (mooring system) 의 연계 해석에 대하여 연구를 수행한다. 그리고 블록 탑재 자동화를 위한 복잡한 장비를 포함한 크레인의 제어 방법을 제안하였다. 블록 탑재 작업의 물리 기반 시뮬레이션 수행 시 필요한 중요한 요소 중 하나는 다물체계 (multibody) 동역학을 기반으로 한 거동 해석이다. 여러 개의 강체 (rigid body) 들이 관절 (joint) 로 연결되어 이루어진 크레인의 운동 해석 뿐만 아니라, 계류 라인과의 완전한 연계 해석을 위해 유연체 (flexible body) 와의 통합된 운동방정식을 구성하는 것이 필요하다. 이를 위해 DELE를 선택하여 1차원 빔 (beam) 과 강체 간의 관절을 비롯한 통합 운동 방정식을 유도하였다. 블록을 탑재하기 위해서 사용되는 크레인은 여러 개의 와이어 로프를 통해 블록을 지탱하며, 해상 크레인의 경우 계류 라인을 통해 위치를 유지하게 된다. 따라서, 탑재 작업 시 일어나는 다양한 현상을 모사하기 위해서는 와이어 로프 및 유연체에 대한 정확하고 효율적인 모델링이 필수적이다. 탑재 시 블록과의 간섭을 고려하기 위해서 구속 조건 기반 와이어 로프 (constraint-based wire rope) 모델을 도입했다. 이를 통해 실제 와이어 로프의 특성인 인장, 구부러짐, 그리고 비틀림을 모델링할 수 있으며, 물체와의 접촉이나 도르래 등과 같은 복잡한 현상도 효율적으로 표현할 수 있다. 계류 라인의 경우, 자체의 동적 거동과 무게에 의한 영향을 고려하기 위해 1차원 빔으로 나타낸 유연체로 모델링하였다. 본 연구에서는 앞에서 제안한 와이어 로프 모델을 통해 작업 중 일어날 수 있는 타 물체와의 상호 작용을 모델링하였다. 블록의 턴오버 작업에서 블록와 와이어 로프 간의 간섭 및 마찰과 같은 상호 작용을 고려하기 위해서 접촉, 슬라이딩, 마찰 모델로 이루어진 알고리즘을 제안하였다. 먼저, 접촉 모델에서는 와이어 로프와 블록 간의 접촉을 탐지하고 접촉 점에서의 힘을 계산한다. 슬라이딩 모델에서는 접촉이 일어난 경우 와이어 로프 방향, 그리고 물체의 엣지 (edge) 방향으로 미끄러지는 거동을 각각 모사하며, 마찰 모델을 통해 정지 마찰력과 운동 마찰력의 영향을 계산한다. 한편, 유연체로 모델링된 계류 라인과 해저면의 접촉은 계류 라인의 위치와 기울기를 제한하는 두 종류의 구속 조건을 통해 모델링하였다. 본 연구에서는 블록 탑재 작업의 크레인 자동 제어를 위한 방법을 제시하였다. 블록의 궤적 추종 제어 (trajectory tracking control) 을 위해 servo constraint라는 개념을 도입하였으며, 이를 복잡한 크레인 시스템에 적용할 수 있는 index reduction 방법을 적용하였다. 그리고 이를 갠트리 크레인과 해상크레인에 적용하여 후크 (hook) 및 이퀄라이저 (equalizer) 가 모두 포함된 복잡한 시스템의 제어를 위한 방법을 제시하였다. 제안한 모델의 유효성을 검증하기 위해 각 모델을 해석해 및 타 상용 소프트웨어의 결과와 비교하였다. 블록과 와이어 로프의 간섭 알고리즘은 각 모델을 해석해 등을 통해 검증하였으며, 계류 라인은 상용 소프트웨어인 OrcaFlex와의 비교를 수행하였다. 그리고 제안한 제어 알고리즘을 통해 블록이 주어진 궤적을 잘 추종하는지 확인하였다. 제안한 모델들은 각각 자체적인 모듈로써 개발이 되었으며, 이는 기존의 다물체계 동역학 기반 해석 프로그램에 통합되었다. 이를 사용하여 블록의 간섭을 고려한 턴오버 작업, 해상크레인과 계류 시스템의 연계 해석을 통한 블록의 리프팅 작업, 그리고 갠트리 크레인과 해상크레인의 자동 제어를 통한 블록 탑재 작업의 해석을 수행하였다. 각각의 해석을 통해 본 연구에서 개발한 모듈들이 성공적으로 적용될 수 있음을 확인하였다. 향후에는 실험이나 HILS 환경 구축 등을 통해 개발한 모델의 추가적인 검증을 수행할 계획이다. The block erection operations in the shipyards generally include block lifting, transporting, turnover, and lowering procedure. During the operation, there are potential risks such as interferences with wire ropes, overweight, collision, which can lead to severe accidents. To check such risks in advance, the physics-based simulations are often required before the actual operation. For this, detailed modeling of the equipment and the operation is necessary to reflect the real situation accurately, which is impossible with conventional modeling methods. Meanwhile, in the actual operation, the position and the orientation of the block should also be precisely controlled to avoid any risks. However, the actual operations are performed manually by the operator, and the cranes used for the erection of the block consist of sophisticated equipment, which makes them difficult to control. In this study, the detailed modeling of the operation situation and the equipment are introduced for accurate simulation. Then, the control method of the crane is suggested for reliable and efficient control. Firstly, the block turnover operation by one floating crane and two crawler cranes is modeled considering the interferences between wire ropes and the block. For this, the interaction model between wire rope and the block, including contact and friction, is suggested. Then, the contact forces exerted on the block are calculated. Secondly, the block lifting by the floating crane is analyzed considering the coupling motion of the floating crane and the mooring system. The interaction among the floating crane, the mooring line, and the seabed is introduced under the various sea condition. Lastly, the control of the gantry crane and the floating crane is performed for block lifting, transporting, and turnover operation. As the gantry crane and the floating crane consist of complicated equipment such as trolleys, booms, equalizers, and wire ropes, the control theory of the underactuated mechanical system is adopted. For the verification, the comparison of the suggested interference model and the analytic solution is conducted. As the proposed model has been divided into three; contact model, friction model, and sliding model, each model is verified independently. Then, for the verification of the suggested mooring line model, the tension and the deflection of the mooring line are compared with that of the commercial software, OrcaFlex. The convergence test was also performed according to the number of the element, to find a proper number of the element of the mooring line. Lastly, the inverse dynamics model formulated in this study is verified by applying it to the dynamic model of gantry crane and check if the block tracks the desired trajectory. To evaluate the effectiveness and applicability of the proposed methods, they were applied to the dynamic analysis of various kinds of block erection operations by using cranes. Three representative applications are provided, such as block turnover operation by floating crane and crawler cranes, block lifting operation by floating crane, and block erection operation by gantry crane and a floating crane. As a result, the proposed method could reflect the actual operation situation accurately, and the block was controlled to the desired trajectory. We conclude that the developed simulation and control methods are applicable to actual operation in ships and offshore structures.

Decomposition-Based Semi-Autonomous Teleoperation Frameworks for Robotic Systems with Distributed Communication and Under-actuation

하창수 서울대학교 대학원 2018 국내박사

The framework of stable bilateral teleoperation has been well established during decades. However, the standard bilateral teleoperation framework could be a baseline for a successful telerobotics but not sufficient for real-application because they usually concentrate on only the bilateral stability. The least considered in the previous research is how to apply a complex robot systems such as multiple mobile robots or a large degree of freedom mobile manipulators for real applications. The main challenges of teleoperation of complex robotic systems in real-world are to achieve two different control objectives (i.e., follow the human command and the coordination/ stabilization of the internal movement) of the slave robots simultaneously, while providing intuitive information about the complicated features of the system. In this thesis, we develop decomposition-based semi-autonomous teleoperation framework for robotic systems which have distributed communication and underactuation property, consisting of three steps: 1) decomposition step, where the human command is defined, and the robotic system is split into the command tracking space and its orthogonal complement (i.e., internal motion); 2) control design of the slave robot, in which we design the slave controller for human command tracking and stabilization/coordination of internal motion space; and 3) feedback interface design, through which we propose a multi-modal feedback interface (for example, visual and haptic) designed with the consideration of the task and the characteristics of the system. Among numerous types of robots, in this thesis, we focus on two types of robotic systems: 1) multiple nonholonomic wheeled mobile robots (WMRs) with distributed communication requirement and 2) manipulator-stage over vertical flexible beam which is under-actuated system. The proposed framework is applied to both case step by step and perform experiments and human subject study to verify/demonstrate the proposed framework for both cases. For distributed WMRs, we consider the scenario that a single user remotely operates a platoon of nonholonomic WMRs that distributively communicate each other in unknown environment. For this, in decomposition step, we utilize nonholonomic passive decomposition to split the platoon kinematics into that of the formation-keeping aspect and the collective tele-driving aspect. Next, in control design step, we design the controls for these two aspects individually and distribute them into each WMR while fully incorporating their nonholonomic constraint and distribution requirement. Finally, in the step of feedback interface design, we also propose a novel predictive display, which, by providing the user with the estimated current and predicted future pose informations of the platoon and future possibility of collision while fully incorporating the uncertainty inherent to the distribution, can significantly enhance the tele-driving performance and easiness of the platoon. The second part is the manipulator-stage over vertical flexible beam which is under-actuated system. Here, the human command defines the desired motion of the end-effector (or the manipulator), and the vibration of the beam should be subdued at the same time. Thus, at the first step, we utilize the passive decomposition to split the dynamics into manipulator motion space and its orthogonal complement, in which we design the control for the suppression of the vibration. For human command tracking, we design the passivity-based control, and, for the suppression of the vibration, we propose two controls: LQR-based control and nonlinear control based on Lyapunov function analysis. Finally, visuo-haptic feedback interface is preliminarily designed for successful peg-in-hole tasks.

캐노피 틸팅 방식 부족 구동 원형 낙하산의 외란관측기 기반 블록 백스테핑 정밀 착지 제어 기법

채수언 경기대학교 대학원 2026 국내석사

This thesis proposes a disturbance observer (DOB)-based block backstepping control mechanism for the precision landing of an underactuated circular parachute employing a canopy-tilting method. A refined nonlinear dynamic model is established, and control inputs are systematically derived by partitioning state variables into blocks and applying a step-by-step backstepping procedure to address the complex nonlinearities of the underactuated system. The primary contribution of this research lies in its novel utilization of the DOB: beyond conventional real-time disturbance compensation, the estimated disturbance data is analyzed to predict the wind direction and velocity within the flight domain. By integrating these wind predictions into trajectory reconfiguration, this approach ensures stable flight and provides a technical foundation for continuously enhancing the mission performance of subsequent flight operations. Numerical simulations demonstrate that the proposed controller achieves robust trajectory tracking and high-precision landing even in the presence of external disturbances. Furthermore, the results validate the potential for progressive performance improvement through the analysis of observation data. 본 논문은 캐노피 틸팅(Canopy Tilting) 방식을 사용하는 부족 구동(Underactuated) 원형 낙하산의 정밀 착지 제어를 위해 외란관측기(DOB) 기반의 블록 백스테핑(Block Backstepping) 제어 기법을 제안하고, 관측 데이터를 활용한 비행 성능 향상 방안을 제시한다. 원형 낙하산 시스템은 구조적 특성상 제어 입력의 수가 자유도보다 적은 부족 구동 시스템이며, 낙하 중 발생하는 불확실한 풍향 및 풍속 등 외부 외란에 매우 취약하다. 본 연구에서는 이를 해결하기 위해 캐노피 틸팅 메커니즘을 반영한 정밀 동역학 모델을 정립하고, 상태 변수를 블록 단위로 분할하여 단계적으로 제어 입력을 도출하는 블록 백스테핑 기법을 적용함으로써 복잡한 비선형 제어 문제를 체계적으로 해결하였다. 특히, 본 연구의 핵심적인 차별점은 외란관측기(DOB)를 단순히 실시간 외란 보상에만 활용하는 것이 아니라, 관측된 외란 데이터를 분석하여 비행 영역의 풍향 및 풍속을 예측하는 데 활용한다는 점이다. 이렇게 예측된 바람 정보는 경로 재설정에 반영됨으로써, 단일 비행의 안정성을 넘어 향후 반복되는 차기 비행의 임무 수행 성능을 지속적으로 향상시킬 수 있는 기틀을 제공한다. 수치 시뮬레이션을 통해 제안한 제어 기법의 유효성을 검증한 결과, 설계된 제어기가 외란이 존재하는 환경에서도 목표 경로를 안정적으로 추종하며 최종 목적지에 높은 정밀도로 착지함을 확인하였다. 또한, 외란 관측 데이터를 통한 비행 성능의 단계적 개선 가능성을 입증하였다. 본 연구의 결과는 향후 자율 낙하산 시스템을 활용한 화물 수송 및 무인기 회수 분야에서 시스템의 신뢰성과 효율성을 극대화하는 데 기여할 것으로 기대된다.

Nonsingular adaptive guidance for impact time and angle control

조동수 서울대학교 대학원 2015 국내박사

The ultimate objective of guided missiles is to intercept targets as small miss distance as possible. Most of the guided missiles employ homing guidance which is a guidance process that can acquire the information of the target with respect to the pursuer and can formulate its own commands to guide itself toward the target. Improvements in missile defense systems pose a great challenge in enhancing the survivability and kill probability of conventional homing guided missiles. Moreover, as manervering capabilities of targets evolve, the more agile and accurate performance of missiles with various constraints are required. This dissertation proposes nonlinear guidance laws whose features are divided into four parts: (i) adaptive guidance, (ii) impact angle control guidance, (iii) impact time control guidance, and (iv) impact time and angle control guidance. First, an adaptive guidance law applicable to a short-range homing missile is investigated against a highly maneuvering target that has the same maximum maneuver acceleration as the missile. In practice, it is difficult to measure the target acceleration with precision and without time delay. The objective of this study is to design a guidance law that can intercept a target within the acceptable miss distance in the presence of unknown target acceleration and guidance command saturation. For this, the main idea of this paper is to apply the fast adaptive control approach to estimating the tangential component of the unknown target acceleration. Moreover, the auxiliary signal is introduced to prevent the presence of the guidance command saturation from destroying the desired adaptive performance. The proposed guidance law can be classied as an augmented proportional navigation guidance law with the actual target acceleration component being replaced by its accurate estimate. Second, a sliding mode guidance law for impact angle control is proposed against a maneuvering target with unknown acceleration, which is capable of achieving the acceptable miss distance and a wide range of the desired impact angle. The main idea is to separate the switching surfaces for the impact angle constraint and the homing constraint, then to associate the two surfaces by introducing an appropriate virtual controller. Because of the unknown target acceleration, an adaptive procedure is designed to select the gain of the switching controller which accounts for the uncertainty bound regarding the target acceleration. The stability of the proposed approach is analyzed by Lyapunov theory, and the capturability analysis is also presented. A guidance problem for impact time control applicable to salvo attacks is considered based on the sliding mode control. To prevent the singularity of the guidance command, a positive continuous nonlinear function of the lead angle is introduced to the guidance command, which makes the Lyapunov stability negative semi-denite. This issue is also resolved by the additional component of the guidance command, which makes the sliding mode the only attractor still without the singularity. The capturability analysis is presented, which is not dependent on the initial launching conditions and can guarantee a wide range of the capture region. The proposed guidance law is easily extended to a non-maneuvering and low-maneuvering target using the predicted interception point. A nonsingular guidance law for impact time and angle control applicable to more eective salvo attacks is proposed by combining the impact angle control guidance and impact time control guidance mentioned above. To incorporate two constraints, the guidance system forms an underactuated system. To deal with this, time-varying slack variables are introduced to form a square system. By introducing a positive continuous nonlinear function of the singular point, the singularity problem of the guidance command is prevented. The presented capturability results are independent on the initial launching conditions of missiles, which can guarantee a wide range of the capture region. The proposed guidance law is easily extended to a non-maneuvering and small maneuvering target using the predicted interception point.

Adaptive Gripper based on Cam-based Underactuated Structure and Shape Memory Polymers

Park,Sungjae 고려대학교 대학원 2025 국내박사

Today, robots are deployed in various environments, expanding beyond traditional industrial sites to the service sector. Industrial robots were initially limited to simple and repetitive tasks such as transporting goods and assembling parts. However, today, robots are increasingly used for complex tasks in various fields including cooking, housework, and medical care, and the ability to perform more complex tasks with a single robot has expanded. As the scope of robot utilization increases, the importance of end effectors, which are robotic tools similar to human hands, is also increasing. Various companies have created various types of grippers such as two-finger parallel grippers, three-finger grippers, and vacuum grippers, and research on robotic hands similar to human hands has also been actively conducted. Despite these developments, the effectiveness of robots in service environments is still limited. The main reason is that various types of grippers are required to perform tasks and handle objects due to limitations in the gripper functions used. For example, two-finger parallel grippers are suitable for standardized objects, while vacuum grippers are optimal for grasping flat and wide objects. Each gripper has a limited work area that it can handle or perform, so it is not suitable for service robots that must handle a variety of objects and perform multiple tasks with a single device. To solve the limitations of these grippers, tool changers have been introduced to provide system diversity by allowing the grippers to be replaced. Tool changers have already been used in industrial settings to facilitate various tasks with a single robot, and have enabled the automation of various tasks. However, the costs associated with tools and tool changers lead to an economic burden because the cost increases with the number of tools used. This cost problem is a fundamental limitation that cannot be easily solved with the concept and configuration of existing tool changers. Robot grippers also require a new concept of grippers. As robots are used in various environments outside of standardized work sites, grippers are required to adapt to unexpected obstacles and collisions with environmental factors and stably grasp objects. Soft grippers have been developed for this purpose, but they are limited in practical use due to limitations such as low durability and weak gripping power. In order to be used in various environments, a special gripper that combines the advantages of soft grippers and rigid grippers with high environmental adaptability, durability, and gripping power is required. In this study, a gripping system consisting of Drivable Automatic Tool Changer (DATC), Cam-based Underdrive Gripper (CUG), and Shape Memory Gripper (SMG) is proposed. The DATC utilizes a mechanism to transmit power from the master of the tool changer to the tool itself, thereby significantly reducing the cost of individual tool components and solving the key cost issues associated with the tool changer. The proposed gripping system can achieve both functionality and versatility by driving multiple grippers with various special functions via the DATC. In order to handle objects of various shapes in environments with various environmental constraints, we propose adaptive CUG and SMG. The CUG has a wide stroke compared to its small size, which enables parallel grasping while minimizing unwanted contact with the environment. In addition, CUG uses a six-bar link structure to implement adaptive gripping based on underactuation that adapts to the shape of various objects, and compensates for vertical and horizontal environmental interactions through a mechanism combining a six-bar link structure and a cam structure. The SMG utilizes shape memory materials to achieve higher adaptability to the shape of the object than the CUG and enables stable adaptive grasping. The SMG achieves adaptability to environmental factors through a leaf spring structure, and unlike the CUG, it enables stable grasping through vertical adaptation without tip rotation. Both the CUG and SMG grippers are special grippers that have the advantages of a soft gripper and a rigid gripper, with high adaptability to the environment, durability, and grasping force. To verify the function of the proposed gripping system, experiments were conducted on DATC, CUG, and SMG. DATC showed a high success rate of 98% in grasping experiments using a two-finger gripper, a three-finger gripper, and a vacuum gripper. In the case of CUG, a high success rate of 82.5% was achieved through vertical adaptation using a scooping motion, and SMG showed a success rate of 82.9% through vertical adaptation. Additional experiments were conducted to confirm the grasping success rates of CUG and SMG with respect to the shape of the object. As a result, CUG was good at grasping flat objects but had a low success rate for cylindrical objects, whereas SMG had a high success rate for cylindrical objects and a low success rate for flat objects. Through this, the two grippers can be interchanged through DATC according to the shape of the object, and a final success rate of 98.3% can be obtained. The proposed gripper CUG, SMG can be combined with a gripping system utilizing DATC to stably handle objects of various shapes in a limited environment. The system has been verified through experiments and has shown high practicality in cases where various tools are required, complex objects are handled, and adaptation to environmental factors is required. 오늘날 로봇은 다양한 환경에 배치되어 전통적인 산업 현장을 넘어 서비스 부문까지 확장되고 있다. 산업용 로봇은 처음에는 물품 운반 및 부품 조립과 같은 단순하고 반복적인 작업에 국한되었다. 그러나 오늘날 로봇은 요리, 집안일, 의료를 포함한 다양한 분야에서 복잡한 작업에 점점 더 많이 사용되고 있으며, 단일 로봇으로 더 복잡한 작업을 수행할 수 있는 기능이 확장되었다. 로봇의 활용범위가 커짐에 따라 인간의 손과 유사한 로봇 도구인 엔드 이펙터의 중요성도 커지고 있다. 다양한 기업들이 2지 평행 그리퍼, 3지 그리퍼, 진공 그리퍼 등 다양한 종류의 그리퍼들을 만들었고, 인간의 손과 유사한 로봇 손에 대한 연구도 활발히 진행되었다. 이러한 발전에도 불구하고 서비스 환경에서 로봇의 효과는 여전히 제한적이다. 주된 이유는 사용하는 그리퍼 기능의 한계로 인해 작업 수행 및 물체를 다루는데 있어서 다양한 종류의 그리퍼가 필요하기 때문이다. 예를 들어, 2지 평행 그리퍼는 표준화된 물체에 적합한 반면 진공 그리퍼는 평평하고 넓은 물체를 잡는데 최적이다. 각각의 그리퍼는 다루거나 수행할 수 있는 작업 영역에 한계가 있으므로, 단일 장치로 다양한 물체를 다루고 여러 작업을 수행해야 하는 서비스 로봇에는 적합하지 않다. 이러한 그리퍼의 한계를 해결하기 위해 그리퍼를 교체할 수 있도록 하여 시스템의 다양성을 제공하는 툴 체인저가 도입되었다. 툴 체인저는 이미 산업 현장에서 단일 로봇으로 다양한 작업을 용이하게 하기 위해 사용되어 왔으며 다양한 작업의 자동화를 가능케 하였다. 그러나 툴과 툴 체인저에 관련된 비용은 사용되는 툴의 수에 따라 비용이 증가하기 때문에 경제적 부담으로 이어진다. 이러한 비용 문제는 기존 툴 체인저의 개념 및 구성에서는 쉽게 해결할 수 없는 근본적인 한계이다. 로봇 그리퍼 또한 새로운 개념의 그리퍼가 필요하다. 로봇의 사용환경이 정형화된 작업 현장에서 벗어나 다양한 환경에서 사용됨에 따라, 그리퍼는 환경에서 예측하지 못한 장애물 및 환경 요소와의 충돌에 대해 적응하고 물체를 안정적으로 파지할 능력이 요구된다. 이러한 목적으로 소프트 그리퍼가 개발되었지만 낮은 내구성 및 약한 파지력 등의 한계로 인해 실사용에 있어서 제한적이다. 다양한 환경에서 활용을 위해서는 높은 환경에 대한 적응성과 내구성 및 파지력을 갖춘, 소프트 그리퍼와 리지드 그리퍼의 장점을 합친 특수한 그리퍼가 필요하다. 본 연구에서는 구동 가능한 자동 툴 체인저(DATC)와 캠 기반 부족 구동 그리퍼(CUG) 및 형상 기억 그리퍼(SMG)로 구성된 그리핑 시스템을 제안한다. DATC는 툴 체인저의 마스터에서 툴 자체로 동력을 전달하는 메커니즘을 활용하여 개별 툴 구성 요소의 비용을 대폭 절감하고 툴 체인저와 관련된 핵심 비용 문제를 해결한다. 제안된 그리핑 시스템은 DATC를 통해 다양한 특수 기능을 갖춘 여러 그리퍼를 구동하여 기능성과 다양성을 모두 얻을 수 있다. 다양한 환경 제약 조건이 있는 환경에서 여러 형상의 물체를 다루기 위해 환경에 적응이 가능한 CUG 및 SMG를 제안한다. CUG는 작은 크기에 비해 넓은 스트로크를 가지며 이를 통해 환경과의 원치 않는 접촉을 최소화하면서 평행 파지가 가능하다. 또한 CUG는 6절 링크 구조를 사용하여 다양한 물체의 형상에 적응하는 부족 구동 기반의 적응형 파지를 구현하고, 6절 링크 구조와 캠 구조를 결합한 메커니즘을 통해 수직 및 수평 환경 상호 작용을 보상한다. SMG는 형상 기억 소재를 활용하여 물체의 형상에 대해 CUG에 비해 더 높은 적응성을 가지고 안정적인 적응형 파지가 가능하다. SMG는 판 스프링 구조를 통해 환경 요소에 대한 적응성을 얻으며, CUG와 달리 팁의 회전 없이 수직 적응을 통해 안정적인 그래스핑이 가능하다. CUG 및 SMG 두 그리퍼 모두 높은 환경에 대한 적응성과 내구성 및 파지력을 갖추어서 소프트 그리퍼와 리지드 그리퍼의 장점을 갖춘 특수한 그리퍼이다. 제안된 그리핑 시스템의 기능을 검증하기 위해 DATC, CUG, SMG에 대해 실험을 진행하였다. DATC는 2지 그리퍼, 3지 그리퍼, 진공 그리퍼를 사용한 파지 실험에서 98%의 높은 성공률을 보였다. CUG의 경우 스쿠핑 동작을 사용한 수직 적응으로 82.5%의 높은 성공률을 달성하였고 SMG는 수직 적응을 통해 82.9%의 성공률을 보였다. 추가 실험을 통해 물체 형상에 대한 CUG 및 SMG의 그래스핑 성공률을 확인한 결과 CUG는 평평한 물체는 잘 잡는 반면 실린더형 물체는 성공률이 낮았고, SMG는 반대로 실린더형 물체의 성공률은 높고 평평한 물체의 성공률은 낮았다. 이를 통해 물체의 형상에 맞춰 DATC를 통해 두 그리퍼를 교체해 사용하여 최종적으로 98.3%의 성공률을 얻을 수 있다. 제안된 그리퍼 CUG, SMG는 DATC를 활용한 그리핑 시스템과 결합하여 다양한 형상의 물체를 제한된 환경에서 안정적으로 다룰 수 있다. 이 시스템은 실험을 통해 검증되었으며 다양한 도구가 필요하고 복잡한 물체를 다루며 환경 요인에 적응이 필요한 경우에 대해 높은 실용성을 보였다.

Terminal sliding surface based finite time control of nonlinear underactuated systems

이신호 Graduate School, Yonsei University 2008 국내석사

Formation control for underactuated autonomous underwater vehicles using approach angle : dynamic surface control approach

김경주 Graduate School, Yonsei University 2012 국내박사