Belief Propagation Decoding with Conflict Search for Polar Codes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing BP based decoding schemes for polar codes face challenges in achieving better performance, low latency, and easy parallel computing in a single decoder, while also dealing with unnecessary computing consumption in updated iterations.
Innovation Solution
The proposed solution is the Belief Propagation Conflict Search List (BPCSL) decoding method, which involves searching for conflict verification processing elements (VPEs) in the BP decoding process to update the right-to-left tables, thereby improving decoding performance and reducing latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If BP based decoding is used for polar codes, then parallel computing capability is improved, but decoding performance is insufficient compared to SC based decoding
Solution Approach 1:
The patent segments the decoding process into multiple independent paths (original path and candidate paths) in the list structure. Each path maintains its own right-to-left table that can be updated independently through parallel conflict verification, allowing the system to explore multiple decoding possibilities simultaneously while maintaining parallel computing advantages.
Solution Approach 2:
The patent performs preliminary conflict verification on right-to-left tables before final decoding decisions are made. By proactively identifying and resolving conflicts in the message passing process, the system prevents performance degradation before it occurs, enabling BP decoding to achieve performance closer to SC-based methods.
2Reliability
If list size is increased in BPL decoding to improve performance, then decoding performance approaches ML bound, but computational complexity and latency increase
Solution Approach 1:
The patent applies conflict verification selectively to specific processing elements (PEs) that show conflicts in their right-to-left tables, rather than uniformly processing all PEs in the list. This localized approach focuses computational resources on problematic areas while leaving other paths unchanged, reducing overall complexity while maintaining performance improvements.
Solution Approach 2:
The patent performs conflict verification only when necessary (when conflicts are detected) rather than always executing full verification procedures. This partial action approach avoids unnecessary computational overhead in cases where the original BP decoding already succeeds, reducing average complexity while preserving performance benefits when needed.
3Reliability
If multiple iterations are performed in BP decoding to improve performance, then decoding accuracy increases, but latency and computational consumption increase
Solution Approach 1:
The patent implements feedback through conflict verification that monitors the decoding process and triggers additional iterations only when conflicts are detected. The conflict verification results feed back into the decoding process, determining whether further iterations are necessary. This feedback mechanism allows the system to stop early when successful while performing additional work only when needed, reducing average latency.
Solution Approach 2:
The patent makes the decoding process dynamic by adaptively determining the number of iterations based on conflict detection. Rather than fixed iterations, the system dynamically adjusts computational effort based on the specific input characteristics and conflict patterns observed during decoding, optimizing the trade-off between accuracy and latency for each decoding instance.
4Reliability
If conflict verification is performed on all PEs in the list, then decoding performance improves, but computational power consumption increases
Solution Approach 1:
The patent applies conflict verification only to PEs that exhibit conflicts in their right-to-left tables, rather than uniformly processing all PEs. This localized verification approach concentrates computational power on problematic PEs while skipping verification for PEs that are already correct, significantly reducing overall power consumption while maintaining performance improvements.
Solution Approach 2:
The patent extracts and identifies only the conflicting PEs from the full list for verification, separating the verification task from the general decoding process. By taking out only the problematic PEs that require attention, the system avoids wasting computational power on PEs that do not need verification, reducing energy consumption while preserving necessary corrections.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Various embodiments of the present disclosure provide methods and apparatuses for improved belief propagation (BP) decoding. A method performed by a receiver comprises: obtaining based on received information and an original left-to-right table comprising left-to-right messages associated with nodes of a plurality of processing elements (PEs) for BP decoding, an original right-to-left table comprising right-to-left messages associated with the nodes, and searching for, based on the original left-to-right table and the original right-to-left table, a conflict verification processing element (VPE) in the plurality of PE. The method also comprises updating the original right-to-left table based on the conflict VPE to obtain a plurality of potential right-to-left tables, and performing the BP decoding based on the respective one of the plurality of potential right-to-left tables.