Korean scientists created robots with memory-mimicking AI that forget outdated data, improving navigation in smart factories and logistics hubs.
DAEGU: A team of Korean researchers has unveiled a breakthrough in robotics by developing artificial intelligence technology that allows robots to mimic human memory — including the ability to forget. This innovation could revolutionise how autonomous mobile robots operate in logistics centres and smart factories by making them more efficient and adaptive.
Developed by the Daegu Gyeongbuk Institute of Science and Technology (DGIST), the new system uses a “Physical AI” model to help robots discard irrelevant or outdated information, much like humans forget unnecessary details. This memory-mimicking approach allows robots to prioritise only current and critical data — such as new obstacles — while forgetting previous obstructions that no longer affect their route.
“By imitating how humans socially forget, we enabled our robots to retain only what matters for efficient navigation,” explained Professor Kyung-Joon Park, lead author of the study published in the Journal of Industrial Information Integration. He added that the work represents a significant step forward in making robotic systems resemble human decision-making behaviour.
Traditionally, autonomous mobile robots (AMRs) face delays when rerouting around temporary obstacles, even after those obstacles are removed. This results in inefficient routing and wasted resources. DGIST’s approach uses collective intelligence and shared memory that evolves in real-time, helping robots avoid unnecessary detours and reduce redundancy in decision-making.
The researchers tested the model in a simulated logistics environment using the Gazebo simulator and found dramatic improvements. Robots using this AI method reduced average driving time by up to 30.1% and improved task completion throughput by up to 18%, compared to conventional ROS 2 (Robot Operating System 2) navigation.
The system relies on standard 2D LiDAR sensors and is available as a plugin for ROS 2, making it highly accessible for industry adoption. The team also suggests potential future uses in autonomous vehicles, drone coordination, smart city traffic systems, and emergency response robots.
This innovation is expected to lead to measurable gains in energy use, operating costs, and equipment durability across various industries.
