Generalized division-free architecture and compact memory structure for resampling in particle filters

Alam, Syed Asad; Gustafsson, Oscar

doi:10.1109/ecctd.2015.7300060

Cited by 1 publication

(6 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The total time required for multinomial resampling is of the order of M 2 as for each iteration, up to M normalized weights must be compared against the random number [24]. However, if the random numbers are ordered, this is reduced to 2M, 1 and, hence, the latency is L r = M.…”

Section: Stratified Multinomialmentioning

confidence: 99%

“…8 from [7]. However, the new expression avoids any division as we can accumulate 1 M to get the 1 M (r − 1) part and the random number generation (RNG) has a range independent of W M . This is easily adaptable for stratified resampling, and, as will be shown later, can be adapted to multinomial resampling.…”

Section: Generalized Division-free Resampling Architecturementioning

confidence: 99%

“…If the number of particles, M, is configurable, it may be required to store multiple different values for 1 M and modify the range of the random number generation accordingly (one way is to multiply a random numberũ ∼ U [0, 1) with 1 M ). Instead, it is possible to rewrite (18) as…”

Section: Generalized Division-free Resampling Architecturementioning

confidence: 99%

“…-Time-update (particle generation/sampling) -Measurement-update (weight computation) -Resampling In the first step, samples (particles) of the unknown state are drawn or sampled from a density using the principle of importance sampling. This step can typically be considered as propagation of particles at time instant n − 1 into n through (1). Thus, this step is also termed time-update.…”

Section: Introductionmentioning

confidence: 99%

“…Preliminary results concerning division free architecture and memory size reduction have been reported in [1]. The rest of the paper is organized as follows: in Sections 2 and 3, basic architectures for particle filters and the resampling process with different types of resampling algorithms are described, respectively.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Improved Particle Filter Resampling Architectures

Alam

Gustafsson

2019

J Sign Process Syst

Self Cite

View full text Add to dashboard Cite

The most challenging aspect of particle filtering hardware implementation is the resampling step. This is because of high latency as it can be only partially executed in parallel with the other steps of particle filtering and has no inherent parallelism inside it. To reduce the latency, an improved resampling architecture is proposed which involves pre-fetching from the weight memory in parallel to the fetching of a value from a random function generator along with architectures for realizing the prefetch technique. This enables a particle filter using M particles with otherwise streaming operation to get new inputs more often than 2M cycles as the previously best approach gives. Results show that a pre-fetch buffer of five values achieves the best area-latency reduction trade-off while on average achieving an 85% reduction in latency for the resampling step leading to a sample time reduction of more than 40%. We also propose a generic division-free architecture for the resampling steps. It also removes the need of explicitly ordering the random values for efficient multinomial resampling implementation. In addition, on-the-fly computation of the cumulative sum of weights is proposed which helps reduce the word length of the particle weight memory. FPGA implementation results show that the memory size is reduced by up to 50%.

show abstract

Section: Stratified Multinomialmentioning

confidence: 99%

Section: Generalized Division-free Resampling Architecturementioning

confidence: 99%