智慧型倒單擺助行車之研發

Abstract

本論文發展一套可作為老人照護機器人之雙輪倒單擺(Two-Wheeled Inverted Pendulum, TWIP)系統。 隨著全世界人口高齡化，銀髮族之行走上的照護逐漸受重視，各種款式的被動型助行器已被廣泛用於幫助老人行走，目前有許多研究人員專注於主動式智慧型輔具之開發，以提供老人更完善的照護。本論文則致力於開發一套具有老人助行、安全機制、及即時步態偵測與評估等功能之雙輪倒單擺系統，此系統具有兩種模式：雙輪模式及助行照護模式。 在雙輪模式之功能實現，我們首先利用牛頓第二定律及拉格朗日法推導系統之數學模型，並以MATLAB SimMechanics作模型驗證。再者將根據理論模型設計強韌控制器來提升系統性能，並在透過MATLAB Simulink模擬後，將其安裝至微處理器進行實驗驗證。實驗過程中我們發現在TWIP系統在平衡控制時，單回授架構並無法克服馬達機構及環境因素所造成的系統干擾，因此進一步提出多回授控制架構來達到穩定的平衡追跡及定位功能。 在助行照護模式時，TWIP系統會由兩輪模式轉換成三輪模式提供老人照護服務，包含助行功能、避障功能、主人追蹤、步態分析及安全機制。在助行時，使用者可藉由握把旁邊的控制盤選擇直行、轉彎或原地休息等模式。此外為了實踐避障及主人追蹤功能，在照護型機器人上方裝置Kinect V2相機，若當主人想獨自行走時，照護機器人將藉由相機之回傳資訊跟隨在主人後方。最後，我們在主人腳踝上配帶穿戴式裝置，以蒐集其步態資料，資料將藉由無線通訊即時傳回Arduino Due控制器，並作即時的訊號分析，且根據訊號分析的結果，進而評估主人當下之步態狀況，若主人步態異常，系統可以予以警示；若主人發生傾倒之瞬間，助行器將即時給予支撐來達到安全上的照護。過程中使用者資料會被妥善保存，以便評估其長期步態之狀況。

This thesis proposes a two-wheeled inverted pendulum (TWIP) system that will be developed as a caring robot for the elderly. As the society is aging, we need to pay more attention to care the seniors. In the walking care, there are many types of passive assistive walkers to help the elder walking. In addition, many researchers have developed active smart walkers to provide better cares. This thesis develops an intelligent inverted pendulum system that combines the functions of walk support, safety design, and real-time gait detection and analysis. The system contains two modes: two-wheeled mode and assistive mode. In the two-wheeled mode, we first derive the model by two methods: the Newton second law and the Lagrange method, and we further verify the derived models by MATLAB SimMechanics. Next, we design robust controllers based on the obtained model. We implement the designed controllers on MATLAB Simulink for simulations, and on the microcontroller for experiments. During the experiments, we find that the single-loop feedback control structure cannot deal with the disturbances caused by the motor mechanism and environment factors. Therefore, we propose a multi-loop control architecture to achieve balancing control and position tracking. In the assistive mode, the TWIP system can transform from the two-wheeled mode to the three-wheeled mode to help the user walk. It provides the following functions: walk support, obstacle avoidance, master following, gait phase detection, and safety design. For walk support, the user can select different functions, such as going straight, turning, and staying, by clicking the control panel near the handle. Furthermore, the robot is equipped with a Kinect V2 camera so that it can avoid obstacles and detect the master. The walker can also follow the master while the user wants to walk himself/herself. Besides, we can attach wearable devices to collect the user’s gait data. The data will be transmitted to the Arduino Due microcontroller through wireless communication for real-time analyses. According to the analyzed results, we can evaluate the user’s condition, and immediately warn the user if the gaits become abnormal. If the system detects the sudden fall of the user, the robots will instantly support and prevent the user from falling. The data will be stored to perform long-term gait evaluation in the future.