| Summary: | 碩士 === 國立清華大學 === 電機工程學系 === 99 === This thesis proposes a particle filter (PF) with threshold Independent Metropolis Hasting(IMH) resampling for the multiple input multiple output(MIMO) fingerprinting
positioning in the cellular system. The proposed technique improves the positioning accuracy in both indoor and outdoor conditions from two perspectives. For the system
perspective, the MIMO system with fingerprinting improves the transmitting diversity and positioning accuracy in the spatial channel model (TR25.996). For the digital signal processing perspective, the proposed particle filter can successfully address the non-linearity issue and combat the non-ideal effect in real channel condition.
The simulation results show that the proposed method has good positioning accuracy, about 5m RMSE when Spacing is equal to 4m. To solve the bottleneck of hardware design in the original particle filter, the thesis proposes threshold IMH resampling algorithm to replace traditional systematic resampling, which has low hardware utilization rate and long processing latency. Further, this thesis proposes selective sampling strategy to reduce the number of particle converging iterations and improve positioning
accuracy.
In the hardware architecture, this thesis designs flexible particle number and sampling threshold according different channel conditions. When the channel condition is bad, we use more particles to estimate the position. On the contrary, we can use fewer particles to reduce the latency. Moreover, using proposed threshold IMH resampling in the hardware, the hardware utilization rate can increase to more than 95% according different particles. Compared to the traditional systematic resampling which needs 3M + L cycles for an iteration, the total processing latency of my design is reduced to M + L cycles, where M is particle number and L is pipeline number.
This study uses ARM 3.2 TSMC-0.18um cell library to implement the particle filter hardware. The operating frequency of the chip is 106MHz when supply voltage is 1.8V.
The core area is 1.48mm2, and the chip area is 3.64mm2.
This thesis anticipates the the implemented particle filter chip is utilized in the base-station to localize mobile users. Hence, the high operating frequency is the major target we want to achieve to make Time-Division Multiple Access(TDMA) scheme practicable in the positioning system.
In the future work, we propose to research implementing our proposed threshold IMH resampling design into PE-CU architecture for further reducing the processing latency
of particle filter.
|
|---|
