Researchers Develop ‘ActionBrain’ that Enables Self-thinking Robots
South Korean researchers have developed a new artificial intelligence (AI) technology that enables robots to complete their assigned missions by making independent decisions on what actions to take.
The state-run Electronics and Telecommunications Research Institute (ETRI) announced on Tuesday that it had developed ‘ActionBrain’ technology that clears the way for robots to make their own decisions on whether to take independent or collaborative actions to carry out their missions.
In Internet of Things (IoT) ecosystems, the way things operate, in general, is based on the rules pre-set by developers.
This way of operation, however, has limits particularly when an unexpected situation occurs or when the environment dramatically changes, with devices having difficulty adapting to new situations.
Against this backdrop, the ETRI developed the ActionBrain technology by applying deep learning technologies such as imitation learning and reinforcement learning.
If this technology is applied to production robots at smart factories, they can develop so-called behavior intelligence for collaborative production, while being able to communicate with other robots.
Even when the actual working conditions are changed from pre-set conditions, the robots can self-correct to ensure optimal operation.
This technology can also be applied to unmanned surveillance drones that can be used to monitor natural disasters or other emergencies.