Amboss demonstrates how machine learning could optimize payment routing on the Bitcoin Lightning Network

A new research paper from Bitcoin Lightning Network payment solutions provider Amboss Technologies on Wednesday outlines how a combination of graph theory and machine learning could be used to enhance the efficiency and scalability of Bitcoin payments on the Lightning Network.

The Lightning Network, recently integrated by Coinbase following other exchanges, is a second-layer solution of payment channels built on top of the Bitcoin blockchain, designed to enable fast and low-cost transactions.

Amboss said that as more crypto exchanges have adopted the Lightning Network, payment reliability has become an increasing challenge. Wallets and exchanges traditionally have used trial and error “probes” to find suitable payment routes across the peer-to-peer network, with Amboss’ machine learning approach seeking to offer a more optimal solution.

“As bitcoin takes center stage in finance, it’s essential that we continue to innovate on Bitcoin scaling technology that further enhances its user experience,” Amboss Technologies CEO Jesse Shrader said in a statement shared with The Block. “Lightning is a critical element to scaling Bitcoin and payment reliability has been its biggest problem. Our research enables scalable, machine learning methods to be an answer.”

Evaluating machine learning models for the Bitcoin Lightning Network

Amboss collaborated with AI firm VantAI and Bitcoin-focused venture capital firm Stillmark on the research, evaluating several machine learning models against traditional heuristic methods, claiming they outperform existing approaches and offer new insights into improved payment pathfinding.

"Our research indicates that machine learning models can predict channel balances with significantly higher accuracy than existing methods," Amboss Technologies Senior Data Scientist Vincent said. "By integrating these models into the network’s pathfinding algorithms, we can streamline operations and reduce payment failure rates, making Lightning an even more robust option for everyday payments."

According to Amboss, the implication is that improved pathfinder efficiency could help broaden the adoption of the Lightning Network, enhancing the utility of Bitcoin as a payment system. The firm now plans to refine the models before integrating them into real-world applications.

In December, Amboss launched “Ghost Addresses,” enabling direct, self-custodial Lightning payments that avoid reliance on custodial wallet providers. Two months earlier, the firm also introduced Hydro, a subscription-based auto-sourcing liquidity solution for the Bitcoin Layer 2.