# Is multiagent deep reinforcement learning the answer or the question? A brief survey

**Authors:** Pablo Hernandez-Leal, Bilal Kartal, Matthew E. Taylor  
**arXiv:** [1810.05587](https://arxiv.org/abs/1810.05587)  
**Date:** 2018-10-12  
**Summary page:** https://www.summarizepaper.com/en/paper/1810.05587v1/

## Key points

- Deep reinforcement learning (DRL) has shown significant progress and achievements in various applications
- Multiagent scenarios have become a key focus area within the field of DRL, with successes in complex multiagent domains
- Challenges still exist in fully leveraging the potential of multiagent deep reinforcement learning (MDRL)
- A recent survey by Pablo Hernandez-Leal, Bilal Kartal, and Matthew E. Taylor provides an overview of current literature on MDRL
- The survey emphasizes successful applications of methods from both DRL and multiagent learning (MAL) in solving problems in MDRL
- The survey aims to distill general lessons from these examples for future research in the multiagent community
- It serves as a bridge between DRL and MAL literature, encouraging collaborative efforts for advancing research in multiagent systems

## Summary

Deep reinforcement learning (DRL) has made significant strides in recent years, showcasing remarkable achievements across various applications and methods. The exploration of multiagent scenarios has emerged as a key focus area within the field, with initial successes reported in complex multiagent domains. However, challenges remain to be addressed in order to fully leverage the potential of multiagent deep reinforcement learning (MDRL). In light of this context, a recent survey conducted by Pablo Hernandez-Leal, Bilal Kartal, and Matthew E. Taylor provides a comprehensive overview of the current literature on MDRL. The survey not only highlights the progress made in this emerging area but also offers valuable guidelines for researchers looking to delve into MDRL. One notable aspect of the survey is its emphasis on showcasing examples where methods and algorithms from both DRL and multiagent learning (MAL) have been successfully applied to solve problems in MDRL. By drawing upon insights from these works, the authors aim to distill general lessons that can inform future research endeavors in the multiagent community. Overall, the survey serves as a unifying force that bridges the gap between DRL and MAL literature, encouraging collaborative efforts towards advancing research in multiagent systems. Through this concerted approach, researchers are poised to tap into a wealth of knowledge and expertise from both domains, paving the way for innovative solutions and breakthroughs in multiagent deep reinforcement learning.