# Multi-agent reinforcement learning for intent-based service assurance in cellular networks

**Authors:** Satheesh K. Perepu, Jean P. Martins, Ricardo Souza S, Kaushik Dey  
**arXiv:** [2208.03740](https://arxiv.org/abs/2208.03740)  
**Date:** 2022-08-07  
**Summary page:** https://www.summarizepaper.com/en/paper/2208.03740v2/

## Key points

- Intent-based management in telecom networks is gaining attention due to performance requirements.
- Traditional closed-loop methods for fulfilling Key Performance Indicators (KPIs) treat each loop independently, leading to performance degradation.
- Scalability is a challenge for existing approaches in intent-based management.
- Techniques like MARL have emerged as a promising solution for coordination and conflict management among loops.
- MARL-based method proposed by Satheesh K. Perepu et al. achieved intent-based management without prior knowledge of the system model.
- The MARL system successfully optimized KPIs for three services in experiments using a network emulator.
- MARL shows potential in enhancing intent-based service assurance within cellular networks, especially with future technologies like 6G.
- Leveraging MARL for intent-based networking offers a dynamic and adaptive framework for efficiently achieving diverse objectives.
- MARL can facilitate interactions between closed loops, prioritize critical KPIs autonomously, and enhance network performance in telecommunications.

## Summary

In the realm of telecom networks, intent-based management has garnered significant attention due to the stringent performance requirements across various use cases. Traditional closed-loop driven methods have been commonly employed in literature to fulfill intents related to Key Performance Indicators (KPIs). However, these methods often treat each closed-loop independently, leading to a degradation in overall performance. Moreover, scalability remains a challenge for existing approaches. To address these limitations,  techniques have emerged as a promising solution for complex coordination and conflict management among loops. In a recent study by Satheesh K. Perepu et al., a method based on MARL was proposed to achieve intent-based management without requiring prior knowledge of the underlying system model. This innovative approach enables MARL agents to incentivize cooperation and promote trade-offs among loops when conflicting intents arise, all without human intervention. The researchers conducted experiments using a network emulator to optimize KPIs for three services. The results demonstrated that the MARL-based system performed admirably, successfully fulfilling all existing intents when sufficient resources were available or prioritizing KPIs effectively in resource-constrained scenarios. This novel approach showcases the potential of MARL in enhancing intent-based service assurance within cellular networks. Looking towards the future, as networks evolve towards technologies like 6G, where network operators' intents will play a pivotal role, leveraging MARL for intent-based networking could prove instrumental. By decomposing intents into objectives and optimizing parameters through closed loops within network slices, MARL offers a dynamic and adaptive framework for achieving diverse objectives efficiently. With its ability to facilitate interactions between closed loops and prioritize critical KPIs autonomously, for enhancing network performance and ensuring seamless service delivery in the ever-evolving landscape of telecommunications.