為了改進現有自走機器人與遠端監控系統在切換不同網域時，必需變更IP位址與網路封包前置碼的缺乏彈性問題、以及當使用者控制機器人時需要攜帶或穿戴感測器才可操控機器人的種種不便；所以本研究將網路服務與立體視覺兩大功能做結合並發展出植基於立體視覺系統與網路服務之自走機器人遠端控制系統，此系統包含以下獨創的設計：(1)一個可移動式結合遠端監控自走機器人平台，此平台結合WiFi無線傳輸並可偵測危險環境中的影像與地形，以及可遠端控制遠端操作之功能，使得此平台具有安全性與機動性的優勢；(2) 一個網路服務資料交換平台，使監控端與設備端可連接至此平台即可隨意設置，使得此平台具有支援不同網域的功能，達到避免重複建置之成本；(3)結合Kinect的立體視覺監控端平台提供使用者不需攜帶感測器或穿戴可辨識之工具，即可將人體的手勢移動結合控制立體視覺偵測轉換成控制命令操作機器人；另外此平台亦可提供監看機器人所處環境情況，才可達到因地適宜的操作。最後本研究對此架構進行整合以及將伺服端部署於雲端系統的效能上做測試比較，驗證各個功能的有效性以及在效能無差異的情況下降低硬體上的建置成本，證實本研究成果可做為新式遠端監控自走機器人之研究參考。
When a robot-car and the remote control system roaming among different network domains, the existing design is to change the IP address and the packet preamble in order to deliver the packet correctly, which is inflexible. Furthermore, the user has to carry or wear sensors in order to manipulate the robot. To improve the inflexibility and the inconvenient shortcomings, we propose a hybrid system that integrates the web service and stereo vision technologies. It includes the following original designs: (1) A robot-car platform with remote control that is capable of communicating with the monitor and control host via WIFI wireless network. In the meantime, it can detect dangerous environments and terrains to ensure the balance of the safety and the mobility. (2) A web service data exchange platform that establishes arbitrary ad hoc connection between the administrator and the robot. It allows the robot-car located in different network domain to carry out communications with the monitor and control platform. As a result, the cost of repeatedly rebuilding the system can be diminished. (3) A controlling host with stereo vision, providing an interface for the users to control the robot-car using merely human body gestures; no extra wearable sensors are required. In addition, it can also be used to monitor the environment of the robot, so as to achieve optimum operation. Finally, we propose an integrated cloud architectures based on the three platforms mentioned above, and compare its efficiency with the local version. The experimental results show that not only can we validate all of the required functionalities, but also can we reduce the hardware construction cost without the system performance trade-off. We believe our results can be a valuable reference for constructing the next generation remote control robot-car.
