An Optimization-Based Parallel Particle Filter for Multitarget Tracking

Abstract

Particle filters are being used in a number of state estimation applications because of their capability to effectively solve nonlinear and non-Gaussian problems. However, they have high computational requirements and this becomes even more so in the case of multi target tracking, where data association is the bottleneck. In order to perform data association and estimation jointly, typically an augmented state vector, whose dimensions depend on the number of targets, is used in particle filters. As the number of targets increases, the corresponding computational load increases exponentially. In this case, parallelization is a possibility for achieving real-time feasibility in large-scale multitarget tracking applications. In this paper, we present an optimization-based scheduling algorithm that minimizes the total computation time for the bus-connected heterogeneous primary-secondary architecture. This scheduler is capable of selecting the optimal number of processors from a large pool of secondary processors and mapping the particles among the selected ones. A new distributed resampling algorithm suitable for parallel computing is also proposed. Furthermore, a less communication intensive parallel implementation of the particle filter without sacrificing tracking accuracy using an efficient load balancing technique, in which optimal particle migration among secondary processors is ensured, is presented. Simulation results demonstrate the tracking effectiveness of the new parallel particle filter and the speedup achieved using parallelization. 111To my mother, who sacrificed so much for my well-being, and to my brother and sister

A New Network for Particle Filtering of Multivariable Nonlinear Objects †

“…The linear complexity provided later resulted in resampling being proposed, e.g., "systematic resampling", which was used by the authors in all experiments presented here. In this resampling, only one random value is needed, and each of the other "drawn" values is higher by 1 N p from the previous one. The systematic resampling is presented in Algorithm 2 (it is assumed that rand() means a random value drawn from uniform distribution U (0, 1)).…”

“…The particle filter (PF) is potentially a very good estimation method because it is based on the optimal solution resulting from Bayesian filtering rules. On the contrary, the biggest disadvantage of PFs is their need for computational power as the number of calculations grows exponentially with the number of system variables [1]. This is the reason why PF methods are usually used only for state estimation with plants of small order.…”

Adaptive Particle Routing in Parallel/Distributed Particle Filters

“…Hong et al (2006) designed and implemented a flexible resampling mechanism for parallel particle filters in a CMOS process, and then analyzed its complexity and performance. Sutharsan et al (2012) presented an optimization-based scheduling algorithm for parallel implementation of particle filters and evaluated the effectiveness of the proposed algorithm by the application of multi-target tracking. Hegyi et al (2007) described two different parallel particle filter algorithms for the state estimation of freeway traffic network.…”

New Grid for Particle Filtering of Multivariable Nonlinear Objects

“…ARTICLE filter (PF) is potentially very good estimation method because is based on the optimal solution -Bayes filter. The biggest disadvantage of PFs is their need for computational power -number of calculations grows exponentially with a system variables number [1]. This is the reason why PF methods are usually used only for very small plants.…”