Distributed Particle Filter with Novel Hybrid Resampling and Balanced Workload

Distributed particle filters (DPFs) enhance the efficiency of particle filtering (PF) for dynamic state estimation by utilizing multiple processing elements (PEs). However, their performance is constrained by the limitations of traditional resampling strategies. Centralized resampling ensures global consistency but incurs high communication costs and computational inefficiencies due to reliance on a single PE. In contrast, decentralized resampling reduces communication overhead but compromises estimation accuracy due to limited local particle diversity. To address these challenges, this paper proposes a novel hybrid resampling algorithm that optimally balances workload among PEs to improve both execution speed and estimation accuracy. Additionally, an FPGA-based architecture compatible with hybrid resampling is introduced to further enhance computational efficiency and processing speed. The proposed design is evaluated using a mobile robot localization task and compared against centralized, decentralized, and existing hybrid resampling approaches. Experimental results show that the proposed method improves estimation accuracy by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$8.5 \%, 9.8 \%$</tex>, and 16.1 % compared to conventional hybrid resampling designs. Moreover, the optimized architecture maximizes resource utilization, achieving a speedup of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1.2 \times$</tex> to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$2.5 \times$</tex> over traditional methods.