Dr. Özcan Dülger, 7 Mart 2018, Çarşamba tarihinde saat 11.30′da Mühendislik Fakültesi seminer salonunda “Uphill Resampling for the Particle Filter and Its Implementation on Graphics Processing Unit” konulu semineri sunacaktır. Seminerin özeti ve kısa biyografi aşağıda verilmiştir.
Uphill Resampling for the Particle Filter and Its Implementation on Graphics Processing Unit
Particle filter has many particles which causes it to have high computational cost. Owing to many cores in its architecture, graphics processing unit (GPU) offers promise for parallel execution of the particle filter. A stage of the particle filter that is particularly challenging to parallelize is resampling because of the interactions among the weights. The most well-known resampling method named, Systematic resampling suffers from numerical instability due to the accumulation of rounding errors when single precision is used. Rounding errors arise in cumulative summation over the weights of the particles when the weights differ widely or the number of particles is large. Furthermore, it is less-readily parallelized because of the interactions among all weights. We introduce a new resampling method, designated Uphill resampling, which is free from numerical instability as it avoids the accumulation of rounding errors. It only compares two weights rather than using the interactions of all weights. It is fast and does not suffer from numerical instability. It achieves similar results with the existing resampling methods on a well-known tracking application.
Özcan Dülger is a research assistant in the Department of Computer Engineering at Middle East Technical University (METU), Ankara, Turkey. He was visiting scholar in Center for Automotive Research at the Ohio State University, Columbus, OH, USA. He received his BS and MS degrees in Computer Engineering from Pamukkale University, Denizli, Turkey in 2007 and 2009 and received his PhD degree in Computer Engineering from Middle East Technical University, Ankara, Turkey in 2018. His research areas are parallel computing, CUDA programming, particle filters and target tracking with radar.