Source localization using particle filtering on FPGA for robotic\n navigation with imprecise binary measurement

摘要

Particle filtering is a recursive Bayesian estimation technique that has\ngained popularity recently for tracking and localization applications. It uses\nMonte Carlo simulation and has proven to be a very reliable technique to model\nnon-Gaussian and non-linear elements of physical systems. Particle filters\noutperform various other traditional filters like Kalman filters in\nnon-Gaussian and non-linear settings due to their non-analytical and\nnon-parametric nature. However, a significant drawback of particle filters is\ntheir computational complexity, which inhibits their use in real-time\napplications with conventional CPU or DSP based implementation schemes. This\npaper proposes a modification to the existing particle filter algorithm and\npresents a highspeed and dedicated hardware architecture. The architecture\nincorporates pipelining and parallelization in the design to reduce execution\ntime considerably. The design is validated for a source localization problem\nwherein we estimate the position of a source in real-time using the particle\nfilter algorithm implemented on hardware. The validation setup relies on an\nUnmanned Ground Vehicle (UGV) with a photodiode housing on top to sense and\nlocalize a light source. We have prototyped the design using Artix-7\nfield-programmable gate array (FPGA), and resource utilization for the proposed\nsystem is presented. Further, we show the execution time and estimation\naccuracy of the high-speed architecture and observe a significant reduction in\ncomputational time. Our implementation of particle filters on FPGA is scalable\nand modular, with a low execution time of about 5.62 us for processing 1024\nparticles and can be deployed for real-time applications.\n