Particle Swarm Optimization-Aided Enhanced Belief Propagation Bit-Flipping Decoding of Polar Codes

Abstract

Belief propagation bit-flipping (BPBF) decoding for polar codes has attracted growing attention owing to its inherent parallelism and improved frame error rate (FER) compared with the original BP decoding. However, existing improved BPBF algorithms still suffer from either a significant performance gap relative to CRC-aided successive cancellation list (CA-SCL) decoding or excessively high computational complexity. Since the constructed bit-flipping set directly determines decoding performance, we propose an enhanced BPBF decoding scheme that employs particle swarm optimization (PSO) to efficiently identify error-prone bits. Specifically, the PSO adopts multiple particles initialized by channel error probabilities to iteratively explore high-risk bits in continuous space. A dynamic inertia weight is introduced to balance global exploration and local exploitation. Breaking the constraints of conventional single-point greedy search, this scheme leverages the globally optimized particle to deduce a reliable bit-flipping combination within the solution space. The simulation results verify that the proposed decoder achieves prominent FER performance gains while maintaining a lower average computational complexity than existing state-of-the-art BPBF decoders based on weighted metric screening.