An accelerated architecture of change‐point detection for FMCW radar mutual interference based on FPGA. (28th August 2021)

• Record Type: Journal Article
• Authors: Yang, Siyuan; Li, Songyi; Wu, Jiayan; Chen, Yong; Liu, Zhenyu
• Is Part Of: International journal of circuit theory and applications. Volume 49:Number 11(2021)
• Page Start: 3719
• Page End: 3732
• Publication Date: 2021-08-28
• Subjects: bottom‐up segmentation -- change‐point detection -- FMCW radar -- FPGA -- mutual interference; Electric circuit analysis -- Periodicals; 621.319205
• DOI: 10.1002/cta.3127
• Languages: English
• ISSNs: 0098-9886

Abstract

Summary: The mutual interference between frequency modulated continuous wave (FMCW) radars seriously affects autonomous driving. The interference will cause a short‐duration pulse‐like signal, which contains several change points in the time domain. The change‐point detection method called bottom‐up segmentation (BOTUP) is applied to deal with a short duration pulse‐like signal, which has excellent detection accuracy. However, the serial processing of BOTUP will cause considerable delay, which is not conducive to real‐time detection of automatic driving. In this paper, an accelerated architecture of BOTUP (ACC‐BOTUP) based on field‐programmable gate array (FPGA) is proposed. As BOTUP has segmented characteristics and without data dependency between the cost functions, a parallel structure is proposed to reduce the latency. Further, a pipeline structure is proposed to execute the operations of the cost functions in an overlapping manner. Compared with the original BOTUP (O‐BOTUP), the latency of ACC‐BOTUP is reduced by 82%, which is more suitable for real‐time detection in autonomous driving. Abstract : An accelerated architecture of change point detection algorithm is designed for multi‐interference in the field of autonomous driving based on FPGA. The accelerated architecture contains two optimization structures: one is a parallel structure and the other is a pipeline structure. This accelerated architecture of the change point detection algorithm is more suitable for real‐time detection in automatic driving.