Research ＆ Product Development

A Complex World Calls for Collaborative Intelligence

In today’s world, a wide range of players—including people, robots, vehicles, equipment, departments, companies, and systems—operate simultaneously and influence one another. As a result, optimizing each component individually can sometimes make the overall system less efficient.
For example, even if one robot selects the shortest route, congestion will occur if every robot takes the same path. Similarly, even if each department optimizes its own operations, the company as a whole may still face inefficiencies in inventory, delivery schedules, and customer service.
What complex environments require is not AI that makes excellent decisions in isolation, but AI that coordinates with others to make better decisions for the system as a whole.
We conduct research and development on this kind of “team-based AI,” or multi-agent AI.

Multi-Agent Deep Reinforcement Learning Library

We are developing a multi-agent deep reinforcement learning (MARL) library that enables multiple AI agents to learn and act collaboratively.

Designed not only for research but also for real-world applications in logistics, energy, manufacturing, mobility, and simulation, it serves as a versatile development platform. Its Python-based architecture makes it easy to use and integrate with existing machine learning environments, enabling engineers to implement and evaluate solutions efficiently.

Through this library, EfficiNet X is expanding the technological foundation for implementing Collaborative AI.

Using the multi-agent deep reinforcement learning library, autonomous trucks, drones, and delivery robots independently determine which packages to carry, creating an efficient logistics network.

Multi-Agent Path Planning Library

We are developing a Multi-Agent Path Finding (MAPF) library that enables multiple robots and vehicles to reach their destinations efficiently while avoiding collisions with one another.

It is a versatile development platform with applications in warehouse transportation, factory logistics, autonomous mobile robots, mobility systems, and traffic simulations.

Its Python-based architecture makes it easy to use and supports route generation and validation in environments containing large numbers of agents and obstacles. Through this library, EfficiNet X is expanding the technological foundation that enables multiple moving agents to coordinate safely and efficiently.

Two thousand robots navigate toward their destinations without colliding with one another.
Our proprietary library enables high-speed route planning for every robot.