الگوريتم مسيريابي تطبيقي مبتني بر يادگيري آنلاين براي پرداخت‌هاي چندمسيره در شبكه لايتنينگ

As the most prominent Layer-2 scaling solution for Bitcoin, the Lightning Network (LN) faces key challenges despite its high potential, including uncertainty in channel capacities, high failure rates for large payments, variable operational costs, an‎d routing delays. This research proposes a novel intelligent routing model that integrates Adaptive Queue Routing (AQR), Multi-Path Payments (MPP), an‎d Online Machine Learning. The objective of this methodology is to enhance payment success rates an‎d optimize costs through dynamic prediction of the networkʹs liquidity state. In the proposed method, each node utilizes a lightweight machine learning model to updat‎e the weights of routing parameters—including capacity, fees, an‎d hop count in real-time based on live experience an‎d previous transaction outcomes. This approach, combined with MPP capabilities, enables the intelligent splitting of payments into optimal shards an‎d their simultaneous routing through parallel paths, significantly reducing deadlocks caused by channel liquidity depletion. To eva‎luate performance, the proposed algorithm was implemented on a custom simulator with dynamic topologies an‎d compared against traditional methods an‎d static multi-path models. Experimental results demonstrate that the proposed model achieves a 94% success rate, representing an improvement of 23% to 32% over baseline methods. Furthermore, the model reduced average transaction fees up to 60% an‎d minimized the number of required retries. Learning curve analysis confirmed that after a short exploration period, the algorithm reaches functional stability an‎d maintains its efficiency under high traffic stress twice as effectively as classical models. While these improvements are accompanied by a controlled increase in end-to-end latency, the balance achieved between success rate an‎d economic efficiency makes this model a viable operational solution for dynamic payment environments.

مسعود عمومي

مجتبي خليلي دليگاني

علي فانيان , الهام محمودزاده