智慧型機器人需要藉助眾多的感測器和制動器來提供豐富的機器人/人互動、機器人/環境互動，為了驅動、控制、監視和資料融合這些裝置，典型的智慧型機器人計算平台需要耗費大量的計算資源，並付出大量功耗代價。本計畫將設計一個機器人智慧感知控制器，用以卸載機器人嵌入式計算平台的主處理器的計算負擔，改善即時性，並減低系統耗電。我們將完成多感測器裝置驅動、資料擷取與基於機器學習的感測融合、裝置管理引擎、和指令控制介面控制器，最後我們將以FPGA完成硬體實作，並分別於四種智慧型機器人(輪式自走車、人型機器人、六軸機械手臂和六足機器人)平台進行軟硬體整合驗證，以展示智慧感知控制器對提升展示機器人控制的即時性、降低耗電與改善系統的可靠度所帶來的好處。 ;Intelligent robots require extensive robot/human interaction, robot/environment interactions with numerous sensors and actuators, thus, typical robotic computing platforms require a significant amount of computational effort to drive, control, monitor, and data fusion of these devices, at the expense of power consumption. The project will design an intelligent perception controller of robot to offload the computational burden from the host processor of the robotic embedded computing platform, improve real-time performance, and reduce system power consumption. We will design and implement the multi-sensors controller, machine learning based sensor fusion controller, device management engine, and instruction interface controller. Finally, we will realize the HW/SW integration verification of intelligent perception controller with four kinds of robot platforms: wheel-type robot, humanoid robot, six-axis robotic arm and hexapod robot, for demonstrating the real-time performance and power consumption reduction.
