Probabilistic gradient descent bit-flipping (PGDBF) is a hard-decision decoder for low-density parity-check (LDPC) codes, which offers a significant improvement in error correction, approaching the performance of soft-information decoders on the binary symmetric channel. However, this outstanding performance is known to come with an augmentation of the decoder complexity, compared to the non-probabilistic gradient descent bit flipping (GDBF), becoming a drawback of this decoder. This paper presents a new approach to implementing PGDBF decoding for quasi-cyclic LDPC (QC-LDPC) codes, based on the so-called variable-node-shift architecture (VNSA). In VNSA-based PGDBF implementations, the regularity of QC-LDPC connection networks is used to cyclically shift the memory of the decoder, leading to the fact that, a variable node (VN) is processed by different computing units during the decoding process. With this modification, the probabilistic effects in VN operations can be produced by implementing different types of processing units, without requirement of a probabilistic signal generator. The VNSA is shown to further improve the decoding performance of the PGDBF, with respect to other hardware implementations reported in the literature, while reducing the complexity below that of the GDBF. The efficiency of the VNSA is proven by ASIC synthesis results and by decoding simulations.

Autonomous vehicles require efficient self-localisation mechanisms and cameras are the most common sensors due to their low cost and rich input. However, the computational intensity of visual localisation varies depending on the environment and requires real-time processing and energy-efficient decision-making. FPGAs provide a solution for prototyping and estimating such energy savings. We propose a distributed solution for implementing a large bio-inspired visual localisation model. The workflow includes (1) an image processing IP that provides pixel information for each visual landmark detected in each captured image, (2) an implementation of N-LOC, a bio-inspired neural architecture, on an FPGA board and (3) a distributed version of N-LOC with evaluation on a single FPGA and a design for use on a multi-FPGA platform. Comparisons with a pure software solution demonstrate that our hardware-based IP implementation yields up to $9\times$ lower latency and $7\times$ higher throughput (frames/second) while maintaining energy efficiency. Our system has a power footprint as low as 2.741 W for the whole system, which is up to 5.5–6× less than what Nvidia Jetson TX2 consumes on average. Our proposed solution offers a promising approach for implementing energy-efficient visual localisation models on FPGA platforms.