LG AI Research Develops 'RAPID', an Industrial Automation Prototyping Platform

Efficient collaboration is essential for robot fleet control, which is core to smart logistics and smart factories. However, despite their high level of sophistication, existing simulation platforms have been difficult to use in on-site environments that require rapid decision-making, due to complex setup and long preparation times.

To solve this problem, LG AI Research developed the RAPID platform. RAPID supports rapid prototyping through the Environment Builder, which lets users intuitively design real-world work environments, and the Simulation Dashboard, which verifies system performance in real time.

The key features of RAPID are as follows:

• Standardized interface: By complying with the PettingZoo API, various MARL (Multi-Agent Reinforcement Learning) algorithms developed by researchers can be applied immediately without any separate code modification.
• Support for diverse behavior models: It supports user-implemented heuristic algorithms, reinforcement learning policies such as MAPPO or QMIX, and LLM (large language model)-based planners such as EXAONE.
• Reproducibility and efficiency: All designs are managed as YAML files for high reproducibility, and performance changes from adding robots or changing the layout can be checked in just a few minutes.

This research was recognized for its practicality and completeness by being selected for the Demonstration Program at AAAI 2026, a globally prestigious AI conference.