智慧型機器人之彈性化即時控制系統

Abstract

本論文以多代理人導向的方式設計一套具有高發展彈性的機器人即時控制系統。本系統的主要目的在於整合異質性機器人控制程式，同時確保系統能有即時的反應效率，並在彈性與通用性之間取得平衡。本架構可幫助開發團隊成員將一複雜之控制系統依照其功能性切割成多個代理人，如此系統的發展得以平行進行，各程式也能共享有限的硬體資源。本系統並提供即時的訊息流通以及決策機制，開發人員只需專注在演算法的開發而不是系統整合的細節部分。在軟體方面本論文使用RTAI這套即時Linux套件來實現，整個架構並提供開放式代理人樣版，使得開發者可以在維持擴充性以及相容性的前提下移植並整合各種異質性的程式。本論文以實驗室研製之家用機器人平台為主，實際製作環境探索、影像追蹤系統及雙機器人合作任務，以驗證本系統的控制效果。

This thesis studies an agent-based approach to developing a flexible real-time robot control system. The main purpose of the system is to integrate heterogeneous algorithms and control methods while still guarantee real-time responding of robotic behaviors. The balance between generality and flexibility is also considered. The proposed architecture facilitates a robot research team to divide a complex control system into agents according to various functions. Therefore, the development can be carried out simultaneously in parallel and integrated in a latter phase. Moreover, the programs can share limited hardware resources onboard the robot. The system also provides a real-time message delivery and decision making mechanism, which help the designers to concentrate on the development of individual algorithm: The software framework of the system is realized using a Linux platform equipped with Real Time Application Interface (RTAI). An agent template is provided with scalability and compatibility to port and integrate heterogeneous programs. Finally, the control system has been tested using a laboratory mobile robot to realize environmental exploration, visual tracking, and a two-robot-cooperation. Experiment results verify the effectiveness of the proposed architecture.