類神經網路應用於機器人軌跡控制

Abstract

在本論文中,我們提出以類神經網路為基礎的控制架構,將其應用於機 器人軌跡控制與追蹤.類神經控制器(Neural Network Controller)學習機 器人動態,結構與非結構之不確定性的適應能力將獲得闡明。經由利亞普 諾夫(Lyapunov)函數的分析我們可以確認控制器的穩定性與收歛。模式學 習(Model learning)也在本論文中使用,模式學習是利用以獲得的動態模 式來對類神經網路做廣義的學習(Generalized learning),其學習方式是 離線的,經由模式學習後的類神經控制器會加快對機器人動態與收歛的學 習速度,模擬結果會顯示此控制方法的可行性與效率。 In this paper, we present a neural-network-based control scheme on thetrajectories tracking for the robotic manipulator. The adaptive capability ofthe neural network controller to learn the dynamics and structured orunstructured uncertainties of the robotic manipulator is demonstrated. Thestability and convergence of the proposed neural-network-based control schemeare guaranteed by the analysis of a Lynapunov function. A model learning isalso used in this thesis. Model learning uses the obtained dynamic model forthe generalized learning of neural networks. The learning procedure is trainedoff line and it is utilized to accerlate learning in the manipulator dynamicsand error convergence with untrained trajectory. Simulations are performedto show the feasibility and effectiveness of the proposed scheme.