Towards rapid dynamic partial reconfiguration in video-based driver assistance systems

Abstract

Using dynamically reconfigurable hardware is useful especially when a high degree of flexibility is demanded and the application requires inherent parallelism to achieve real-time constraints. Depending on various driving conditions different algorithms have to be used for video processing. These different algorithms require different hardware accelerator engines, which are loaded into the AutoVision chip at run-time of the system. The novelties presented in this paper are the determination of the maximum frequency for dynamic partial reconfiguration of Xilinx Virtex-II Pro, Virtex-4 and Virtex-5 devices and a modified overclocked version of the ICAP controller. In addition an online verification approach is presented that can determine configuration errors that might be caused by configuring a device above the specified frequencies. This results in a reconfiguration throughput which is three times higher than the maximum throughput specified by Xilinx. © 2010 Springer-Verlag.