Positioning Particle Filter Using MIMO Fingerprinting for Cellular Communication System

適用於行動通訊系統下使用多輸入多輸出特徵定位之粒子濾波器

鄭旻欣(Min-Hsin Cheng)

指導教授 ： 黃元豪

國立清華大學/電機資訊學院/通訊工程研究所/碩士(2010年)

摘要

本篇論文提出適用於行用通訊系統下多輸入多輸出特徵定位之粒子濾波器。我們提出的技術從兩個觀點來改善在室內與室外的定位精準度。以系統的觀點，於空間通道模型(TR25.996)下使用多輸入多輸出系統特徵增進定位精準度。以訊號處理的觀點，粒子濾波器可以成功地對抗非線性問題。應用於2x2多輸入多輸出系統，我們提出的多輸入多輸出特徵定位之粒子濾波器比傳統的多輸入多輸出粒子濾波器減少約66%乘法運算。就我們所知，唯一假設通道資訊非完美已知僅有我們所提出的多輸入多輸出特徵定位之粒子濾波器。模擬結果顯示多輸入多輸出特徵定位之粒子濾波器有很好的定位精準度，於Spacing為4公尺情形下平均時間的均方根誤差約為5公尺。

對於硬體架構方面，我們提出兩種方法加速現有的處理單元-核心單元(PE-CU)架構並且減少粒子間的交換。首先，PE使用臨限取樣/權重(threshold sampling/weighting)藉由忽略小於界限的粒子運算來降低處理時間。再者，在沒有粒子間交換的情形下平均權重核心單元(weight-balanced CU)避免了效能下降。除此之外，我們應用Independent Metropolis Hasting重新取樣(IMH resampling)管線化處理重新取樣步驟。模擬結果顯示我們提出的架構相較於最大概似取樣-區域重新取樣(ML-LR) PE架構不僅提升了14.8倍的速度，同時均方根誤差也優於ML-LE PE架構約9公尺當PE個數為2。總結此論文，在通道資訊並非完美已知的情形下多輸入多輸出系統特徵定位粒子濾波器增進了精準度，同時我們提出的架構也改善了速度與精準度。

關鍵字

粒子濾波器 ； 特徵定位 ； 多輸入多輸出系統 ； 行動通訊系統 ； 3GPP-LTE ； 幾何通道模型

並列摘要

This thesis proposes the multiple input multiple output (MIMO) fingerprinting positioning particle filter (PF) for cellular system.
The proposed technique improves the positioning accuracy both indoor and outdoor from two perspectives.
For the system perspective, the MIMO system with fingerprinting improves the accuracy in the spatial channel model (TR25.996).
For the digital signal processing perspective, particle filter can successfully address the non-linearity issue.
For the 2x2 MIMO system, the proposed MIMO fingerprinting positioning particle filter can reduce about 66% multiplications of the traditional MIMO positioning particle filter in [1]. To our best knowledge, the proposed MIMO fingerprinting particle filter is the only one based on the assumption that the channel information is not perfectly known.
The simulation results show that the proposed method has the good positioning accuracy, an over-time mean RMSE about 5 m when
$Spacing$ is equal to 4m.

In the hardware architecture, we propose two techniques to speed up the current processing element-central unit (PE-CU) architecture and reduce the particle communications.
First, PE uses the threshold sampling/weighting to reduce the latency by passing over the computation of particles having difference smaller than threshold.
Second, weight-balanced CU avoids the performance degradation under the situation of no particle communications.
Moreover, we pipeline the resampling step by applying IMH resampling.
The simulation results show that the proposed architecture not only speeds up to 14.8 times faster than ML-LR PE architecture
but also achieves better RMSE about 9m than maximum likelihood prior sampling-localized resampling (ML-LR) PE architecture when the number of PEs is two.
To conclude this work, the proposed MIMO fingerprinting positioning particle filter improves accuracy based on the assumption that the channel information is not perfectly known in the spatial channel model,
and the proposed architecture improves both accuracy and latency.

並列關鍵字

particle filter ； fingerprinting positioning ； MIMO system ； cellular communication system ； 3GPP-LTE ； Spatial channel model

參考文獻

[5] Spatial channel model for Multiple Input Multiple Output(MIMO) simulations(Rel.7), 3GPP Std. TR 25.996, Rev. 7.0.0, Jun. 2007.

[1] S. Bizjajeva, T. Ryden, and O. Edfors, “Mobile Positioning in MIMO System Using Particle Filtering,” in Proc. IEEE 66th Veh. Technol. Conference, Sep. 2007,pp.792–798.

[2] P. D. et al., “Particle Filtering,” in Proc. IEEE Signal Processing Magazine, Sep.2003, pp. 19–38.

(SiPS-2010), San Francisco, USA, Oct. 2010, p. 1-4.

[6] A. Sankaranarayanan, R. Chellappa, and A. Srivastava, “Algorithmic and architectural design methodology for particle filters in hardware,” in Proc. Int. Conf.Computer Design, Oct. 2005, pp. 275–280.