AI BriefingLG AI Research: Multi-Agent Reinforcement Learning (MARL) Research for Solving Real-World Problems
Key point
LG AI Research presented multi-agent reinforcement learning research for autonomous driving and industrial site optimization at AAMAS 2024.
Details
LG AI Research's Data Intelligence (DI) Lab presented two papers applying multi-agent reinforcement learning (MARL) to real-world problems at the international conference AAMAS 2024. This research focused on agent modeling and algorithm advancement for applying AI technology to industrial sites.
The first study, 'Surge Routing', addresses an On-demand Mobility routing algorithm for autonomous taxi services. To overcome the limitation of existing algorithms failing to handle sudden demand changes caused by large-scale events, the researchers proposed a new framework combining a demand forecasting model that leverages event data with MARL.
This framework operates with the following structure:
- Event processing module: Collects and processes data such as event titles, descriptions, and reviews from the internet
- Demand forecasting module: Predicts demand at the time and location of events
- MARL framework: Performs efficient routing in a city-scale environment using the 'One-agent-at-a-time Rollout' method based on the forecasting results
The second study is a paper by LG AI Research introducing research that uses MARL to optimize scheduling for a Naphtha Cracking Center.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.