以Particle Filter實現分散式救援機器人之ZigBee訊號定位系統

Abstract

本研究完成以TI DSP2812為核心之四軸馬達控制電路，以此電路整合感測、通訊以及運動控制介面，透過此驅動電路快速建立整個救援機器人硬體系統，並且設計遠端監控系統監控機器人感測資訊。此外透過此驅動電路版，可以讓有心想要接觸機器人領域的初學者，藉由設計過的實驗課程，很快的了解機器人系統的硬體建立方式與感測器的運用，最後建立屬於自己的機器人平台。 本研究研發的救援機器人搭載透過以網狀拓樸結構的ZigBee網路無線介面通訊，除了通訊功能之外，附帶有接受訊號強度指標（Receiver Siganl Strength Index），可以將訊號強度換算成距離，只要有兩個參考點，就可以利用Trilateration達到絕對定位的目標，但是以無線訊號強度定位的方式容易受到多重路徑與天線方向的影響，導致無線訊號強度產生誤差，導致定位誤差。另外目標物位置與參考點的相對位置，也會對定位的精準度產生影響，本研究以Particle filter融合機器人感測資訊，減低多重路徑與天線方向對定位造成的影響，並且與Kalman filter 比較定位精準度，與未補償之定位結果相比定位誤差改善73.9%，最後以繞行一方型的方式呈現定位的效果，與未補償之定位結果相比，能夠清楚看出機器人移動軌跡。

The 4-axis motion control card is designed in this thesis with the TI DSP2812 microcontroller to provide all motion control, sensor capture, and wireless communication for the rescue robot. With the 802.15.4 ZigBee wireless sensor network system, a built-in received signal strength indicator (RSSI) is adopted to estimate the location of the target by applying the Trilateration algorithm. Since the RSSI is affected heavily by the multi-path effect and antenna orientation which will cause serious localization error, the particle filter is adopted in this thesis with the fusion technique to significantly reduce the effects of multi-path and antenna orientation. Experiment results indicate that the localization accuracy applying the proposed particle filter fusion technique is greatly improved 73.9% compared with applying the Kalman filter only. Experimental results for the robot motion with a square contouring indicate that the proposed method can reliably estimate the real robot trajectory based simply on the RSSI of the ZigBee system.