A research team at the Daegu Gyeongbuk Institute of Science and Technology (DGIST) has developed a method for improving the efficiency of autonomous mobile robot (AMR) navigation by incorporating human-like information processing into their operation. Led by Professor Kyung-Joon Park of the Department of Electrical Engineering and Computer Science and the Physical AI Center, the team introduced a “Physical AI” technology that mimics how social information is spread and forgotten over time.
The research addresses a common challenge in logistics and manufacturing environments where AMRs frequently encounter unpredictable obstacles such as forklifts and misplaced cargo. Traditional systems rely on reactive behavior, which can result in inefficient detours and delays. In contrast, the new approach enables robots to share essential information while discarding irrelevant data over time, allowing for more coordinated and efficient navigation.
The technology was tested using the Gazebo simulator, a virtual logistics center environment. Results showed that robots using the system completed tasks up to 18% faster and reduced average driving time by as much as 30.1% compared to conventional navigation methods based on ROS 2.
According to DGIST, the system can be implemented using 2D LiDAR without additional hardware, and it functions as a plugin compatible with the ROS 2 navigation stack. This compatibility is intended to facilitate integration into existing AMR systems across applications including logistics, drone coordination, autonomous vehicles, and potentially smart city traffic management and large-scale rescue operations.
