Region Guided Shortest Path Algorithm for Dynamic Networks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Classic shortest path algorithms are inefficient for large graphs, leading to impractical computation times in communication and social networking applications, where real-time responses are required, especially in dynamic network topologies.

Innovation Solution

A two-pass random N-seeding clustering algorithm partitions network nodes into regions, precomputes intervals between regions, and uses region-guided search to determine shortest paths efficiently, reducing unnecessary computation and adapting to network changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If classic shortest path algorithms (Dijkstra, Bellman-Ford, Floyd-Warshall) are used on large graphs, then computation accuracy is maintained, but computation time becomes impractical and cannot meet real-time requirements

Engineering Contradiction:
Improveshortest path computation accuracyVSAvoidcomputation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent partitions the graph into multiple clusters, where each cluster contains a subset of vertices. This segmentation allows the shortest path computation to be divided into two stages: first computing paths within individual clusters, then combining results across clusters. This reduces the computational complexity from O(V³) for the entire graph to O(k·V²) where k is the number of clusters, significantly speeding up computation while maintaining accuracy through the two-pass algorithm that first computes intra-cluster shortest paths and then inter-cluster paths.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary computation of shortest paths from a source vertex to all vertices within its own cluster before attempting to reach vertices in other clusters. This preliminary action allows the algorithm to prune the search space by eliminating vertices in other clusters that cannot be reached more efficiently through the current cluster, thereby reducing computation time while preserving the ability to find the true shortest path.

Inventive Principle:
Principle #10Preliminary action

2Loss of time

If graph preprocessing is performed to speed up shortest path queries, then query response time improves, but the system complexity increases and adaptability to dynamic network changes decreases

Engineering Contradiction:
Improvequery response timeVSAvoidsystem complexity
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent creates a clustered structure that segments the graph into manageable regions with identified border vertices. This segmentation serves as preprocessing that speeds up queries by limiting the search to relevant clusters and their borders, while the modular nature of the clustering makes the system adaptable to changes. When network topology changes, only local cluster structures need adjustment rather than reprocessing the entire graph.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent employs a dynamic two-pass algorithm that adapts to changing network conditions. The first pass computes shortest paths from the source to border vertices of its cluster, and the second pass uses this information to prune the search space for reaching other clusters. This dynamic approach allows the algorithm to adjust its computation based on real-time network state, maintaining both speed and adaptability without requiring complete reprocessing when changes occur.

Inventive Principle:
Principle #15Dynamics

3Reliability

If the entire graph is processed to ensure complete path coverage, then path finding completeness is guaranteed, but computation resources are wasted on vertices that cannot be reached

Engineering Contradiction:
Improvepath finding completenessVSAvoidcomputation resources
Core Design Contradiction:
ReliabilityVSLoss of energy

Solution Approach 1:

The patent performs a preliminary computation of shortest paths from the source vertex to all vertices within its own cluster before attempting to reach vertices in other clusters. This preliminary action enables the algorithm to identify which vertices in other clusters are actually reachable and worthwhile to process, allowing it to prune the search space and eliminate computation on unreachable vertices while still guaranteeing completeness for all reachable vertices.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

By segmenting the graph into clusters with identified border vertices, the patent enables targeted processing where only relevant clusters need to be explored. The algorithm can determine early whether certain clusters are reachable from the source and focus computational resources only on those clusters, ensuring complete path finding for reachable vertices while avoiding waste on unreachable portions of the graph.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3295616B1Region guided and change tolerant fast shortest path algorithm and graph preprocessing framework
Publication Date: 2020.05.06 HUAWEI TECH CO LTD
  • EP3295616B1 patent drawingFigure 1
  • EP3295616B1 patent drawingFigure 2~3
  • EP3295616B1 patent drawingFigure 4~5

AI summary

A method for region guided and change tolerant fast shortest path determination and graph preprocessing for network management and control. In an embodiment, a method includes partitioning, by a network component, a plurality of network nodes into a plurality of regions, each network node belonging to one of the regions; identifying, by the network component, border nodes for each region, each border node in a region connecting to at least one border node in a connecting region; determining, by the network component, intervals between regions according to the border nodes, each interval comprising a minimum distance and a maximum distance between two regions; determining, by the network component, a path from a source node to a target node according to the intervals.